Side Hustles for Data Scientists

00:00 Are you a data scientist looking to branch out on your own and start something new?

00:03 Maybe you're just looking for a way to work with those exciting libraries that aren't yet

00:07 in play at the day job. Rather than putting everything on the line, quitting your job,

00:12 and hoping things work out, maybe you should start a side hustle. On this episode, you'll

00:16 meet Keith McCormick, a data scientist who has many irons in the fire, and he's here to tell us

00:20 about the different types of side hustles and why you might want to try or avoid a certain one.

00:25 This is Talk Python to Me, episode 290, recorded October 1st, 2020.

00:30 Welcome to Talk Python to Me, a weekly podcast on Python, the language, the libraries, the ecosystem,

00:49 and the personalities. This is your host, Michael Kennedy. Follow me on Twitter where I'm at

00:53 mkennedy and keep up with the show and listen to past episodes at talkpython.fm and follow the

00:58 show on Twitter via at Talk Python. This episode is brought to you by Linode and Talk Python Training.

01:05 Please check out the offers during their segments. It really helps support the show.

01:08 I got a joke for you. What's the world's most popular IDE? Excel. Funny, right? Except many

01:15 companies really do run on Excel to the point where they would be much better off using clean and simple

01:21 programming tools. For many, Python's data science stack would be vastly better. But moving from Excel

01:27 to Python is a challenge. Most data science courses don't focus specifically on the Excel use cases.

01:32 That's why we've teamed up with Chris Moffitt from Practical Business Python to create a course

01:37 tailor-made for helping people learn just enough pandas in Jupyter to replace the problematic Excel

01:43 usage with clean and scalable Python code. If you or your coworkers are ready to move beyond Excel,

01:49 visit talkpython.fm/Excel, or just click the link in the show notes to learn more about this

01:54 online course at Talk Python Training. Keith, welcome to Talk Python To Me.

01:58 Thanks very much. I look forward to it.

02:00 I'm really excited about this topic. As folks who listen to the show know, I'm very excited about

02:05 entrepreneurship, about software businesses, businesses built on open source. And so many of

02:12 these things often I think they seem inaccessible to folks because, well, they've got a mortgage to pay,

02:17 or they've already got a job, or they don't have a chance to just go through some Y Combinator,

02:24 accelerator type experience, and then pop out with VC money on the other side. And we often hear about

02:30 people creating side hustles. And for some reason, maybe it's just the world I live in, the water I swim in.

02:35 But to me, I hear that so often on the software development side, especially the front end

02:40 development side, but not on the data science side so much. So that's what we're going to talk

02:45 about today is side hustles for data scientists.

02:47 Yeah, I'm excited. When I reflect back, I've never really thought that I had any special

02:52 skills or knowledge. But just within the last couple of years, looking back, I said, well, gosh,

02:57 I've been doing this on my own for 25 years. Only briefly was I full time. I must have learned

03:05 something during that time. So I never thought I had a story to share. But I guess if you do it this

03:09 many years, I've definitely picked up some things.

03:11 I would think so. And congratulations. That's awesome. You've definitely cracked the code and

03:16 made it work if you can do it for 25 years.

03:17 I was just going to say that depending on how you count, I guess it's been longer than that,

03:22 because I finished undergrad in 91 and I've only had one W-2 style job. I was an army scholarship. I

03:30 guess that counts as a traditional job. But I ran a tutoring company. I wasn't a data scientist yet,

03:36 but I ran a tutoring company because I had done some SAT prep, you know, the college prep stuff after.

03:41 So maybe that's how I got the entrepreneurial bug. I don't know. But if you start counting from there,

03:46 even though I wasn't doing data science as an entrepreneur, I guess we're at 30 years.

03:52 Yeah. Yeah. Well, I was going to say, I'm a relative rookie. I've only been doing this for

03:56 five years on my own.

03:57 And I think the market is probably different. I mean, who knows what it would be like if I were

04:02 to start the clock all over again. So that will probably be part of what we can uncover is how

04:06 much.

04:06 Yeah, I totally agree. And I would say it's easier and it's harder. It's easier because it's easier to

04:12 get the word out. It's cheaper, right? We have all of these async communication and collaboration

04:18 things. We've got GitHub, we've got Zoom, and yet there are now millions of people vying for that

04:25 same attention. So getting noticed, I think, is actually harder. So there's like this, you're more

04:29 capable of getting the word out, but there's because of that so many people trying. So I think that

04:33 it's still challenging, but in a different way somewhat.

04:37 You know, I think you're exactly right. Back in the nineties, if you had a conversation about this

04:41 kind of stuff, potential client would say, you know about that? We need to find somebody that

04:47 knows something about that. It's not like there was a lot of competition. You were the person they knew

04:51 that knew about this stuff.

04:53 Yeah. Yeah. You could write a book and then that would get you, well, that's the expert. They wrote

04:57 the book. All right. Now it's like, it's not necessarily going to do it. All right. So I'm super

05:02 excited to dive into all these things, but let's start with your story. How did you get into

05:06 programming and how do you get it? I know you're not exclusively or even primarily doing Python in

05:11 your data science work, but let's talk a little bit about how you got into Python anyway.

05:15 Sure. Well, perhaps I should mention how I got into data science and then pick up the programming

05:21 piece. Sure. So the way that I got into data science was the way that most of us did who got

05:27 into this in the nineties. Back then it was expected that you had a substantial statistics background.

05:35 Now I think you can get away without that, but at the time that was just table stakes. You just had

05:41 to have a stats background. So I somewhat stumbled into statistics. I studied computer science as an

05:47 undergrad, not statistics, but I was contemplating grad school and I had to do something to pay the

05:54 bills. So I was looking for ways to do that. That wouldn't be full-time. And I got an opportunity

05:58 to teach introductory classes for SPSS, the company that made the famous statistics software.

06:06 Right. This is kind of like MATLAB, but it was focused. It's a tool focused on statistics explicitly.

06:12 And it was really popular in the nineties. Yeah.

06:14 Yeah. And boy, I mean, it's just been around forever. So it actually, I'm in my early fifties.

06:20 I was born in 68 and that's actually the year that SPSS came out. So I'm SPSS and myself were the same

06:26 age. And SAS, which of course is also famous, came out just a couple of years later, 71 or something like

06:32 that. So they've both been around for a half century. It's really remarkable. And you're right. In the

06:36 nineties, they largely dominated the statistics space, but they also completely and utterly dominated

06:43 the predictive analytics space because at the time in the late nineties, they were the only game in town.

06:48 So in SPSS, it was a product called Clementine, which is now an IBM product. So it now is called

06:56 a somewhat odd name, IBM SPSS modeler. And SAS, shortly after Clementine came out and made something

07:04 called enterprise minor. And those were really the tools that people used if they were interested in

07:08 this kind of stuff. So I backed into Python. I mean, I had studied computer sciences in undergrad. So

07:14 certainly the notion of coding and thinking programmatically made a lot of sense to me,

07:20 but SPSS Inc for whatever series of reasons chose to make Python, the scripting language in both SPSS

07:29 statistics and SPSS modeler. So we don't think of those tools. If folks are familiar with them,

07:35 I don't think of those tools as having a programming component, but just like Excel has macros,

07:40 these tools have considerable programming options as well. And they decided that the languages that

07:47 were around in 68 weren't exactly cutting edge anymore.

07:51 Cobalt's just not doing it. People don't love it for some reason.

07:56 For a time, organizations were desperate for COBOL programmers, all of which were basically

08:02 semi-retired, but would be pulled out of retirement to get crazy hourly fees to fix ancient COBOL code

08:08 that I'm not quite old enough to be a COBOL person. But yeah, the equivalent, you know, like SPSS

08:14 modeler initially had a kind of a prologue type, that more 90s AI language type stuff.

08:22 Oh, wow. Yeah. Okay.

08:23 Stuff like that was the basis of how you would do some basic scripting and modeler. So both SPSS

08:28 statistics and SPSS modeler needed to be renovated in terms of their scripting language. And Python was

08:34 the one that was adopted. So I sought out experts on Python to mentor me. And I took a class with Mark

08:41 Lutz, which some people might know because he wrote this big doorstopping comprehensive encyclopedic

08:47 O'Reilly book on the subject. And then not only- Yeah, this is a massive book. It was a good book.

08:52 It is a good book, especially at the time. I think this was like 2005 or so. So in a sense,

08:57 I was early to the party. And in another sense, I'm still late to the party in that I've been using

09:03 these tools for so many years. I'm definitely not a Python all the time kind of a guy.

09:08 Excellent. Yeah, that's really cool. And it sort of SPSS sort of dragged you into it. That's cool.

09:13 So how about now? I always like to ask people what they're doing day to day. So folks listening,

09:19 get a sense of where you're coming from. How do you spend your time and stuff?

09:22 Sure. No. Well, let's just kind of finish up on the tool thing because folks might,

09:27 after having heard all that, they might be kind of curious, kind of what my GoTool generally is.

09:32 So the first book that I wrote was on SPSS Modeler. So I'm well-known in that community

09:40 because it wasn't just my book. It was the first book on that subject. And I gathered all of us that

09:47 were well-known in that community and ended up being this group thing. And I was lead author. So

09:52 I'll always be well-known in that community, but that community is shrinking in importance relative

09:56 to others because it's no longer one of two options like it was 20 years ago, clearly. And it's not the

10:03 dominant option because Python is the dominant option. So I'm known for that, but it's not a big part of my

10:08 consulting anymore. So the tool that I like to use, if I'm just doing a workshop where I'm working with a

10:16 client that's really just starting out and we don't want to be distracted by the tool, we just want to be able to

10:21 talk about the project is nine. And that's a predictive analytics workbench that's open source

10:27 that obviously isn't nearly as known as Python, but has kind of a growing following. And the only reason

10:31 I use that is that if I'm working with a client, I don't want to be focused on the project within an

10:36 hour or so we can be doing very basic stuff. And then the tool fades into the background, which I like.

10:41 But then for all the other clients, I just use whatever they use. And I find that I can be helpful to

10:46 them without having to worry about that.

10:49 Right. Probably a lot of Jupyter, some R, some Python.

10:51 All of the above.

10:52 Yep.

10:53 Yep. All right. Day to day. How do you spend your time?

10:56 Well, in some ways it's easier to talk about how I kind of carve up my year because let's say I'm

11:02 going to speak at a conference.

11:03 It's not like you just work for one company 50 weeks a year, then you got two weeks off, right?

11:08 You've got probably, as we're going to talk about, a mix of things you're doing for different

11:12 companies and different people, right?

11:13 Absolutely. Yeah. So it's not even just...

11:16 So if, for instance, if one week I was going to work for client A and then the next week I was

11:20 going to work for client B, then you would get a sense that my day to day might be somewhat more

11:24 like someone that was on salary. But my year is all over the place because the clients that work

11:30 with me, generally speaking, I'm not building the predictive model. I'm usually helping them run

11:35 their teams. And the reason I've gravitated towards that is it feels right at this stage in my career,

11:40 but also because it fits my schedule better. Because if I'm going to give a talk and a workshop

11:46 in Amsterdam or something, I want to be able to tune out and not worry about being up in the middle of

11:52 the night on some client meeting or something like that. I'll tell my clients that I'm out of pocket

11:56 for a week or 10 days. And to be honest, if I couldn't do that, it probably wouldn't be fun because

12:01 these long flights, you get tired and you want to enjoy the fact that you're in Europe or you want

12:04 to enjoy the fact that you're in Kuala Lumpur or what have you. So the weeks that I'm at a conference,

12:09 I'm all about the conference. And I'm really just giving an out of office reply for email,

12:14 but my clients will know that I'm going to be out of pocket for a week or two. So that's maybe 20%

12:21 of my year is that kind of travel. Oh, the other thing that I'll do that some of my colleagues don't

12:26 have the time to do depends on if they have kids and the age of their kids and different things like

12:31 that. But a lot of my colleagues have to run home to catch the soccer game or whatever.

12:35 I'm single without kids, so that's not an issue for me. So I'll tend to stay traveling. And that

12:42 really becomes a really fun part of it for me. Yeah. So that's a substantial chunk. And then most of the

12:49 consulting that I'll do is more like on a retainer basis. So I might have four or five hours with one

12:57 client and eight with another combined with other things. And of course, the piece that if folks

13:02 haven't done a lot of freelancing, they have to be prepared for is you have to be interacting with

13:08 new clients, scoping projects, you know, maybe sending out invoices or project proposals.

13:15 And that's a constant thing. If you're not doing that all the time, you're going to suddenly reach

13:20 the end of a project and realize that you have nothing going on. So that's constantly happening.

13:25 And it's something that it's hard to get prepared for. How do you get started doing that? You know,

13:29 it's one thing to say, okay, I've been working at this company for three years. I'm really good at

13:34 X, you know, building APIs, doing either doing this type of data analysis, whatever. But that's only half

13:40 of the story. Like if you want to go into consulting or go into speaking and you've got to build these

13:46 relationships and build that. It's almost like a funnel, but for your career instead of for a product.

13:52 Well, absolutely. It's very much a funnel. I think part of it is, is that at a certain point in my career,

13:59 I would have this kind of boom or bust situation, which is quite common when people try to consult.

14:06 But it's especially true when you're doing a 40 hour week for one client. So at this stage of my career,

14:13 I never do that anymore. I would have had to accept those kinds of gigs years ago to pay the bills.

14:18 But I've been kind of graduated past that, so to speak, because if you are working with one client,

14:24 40 hours a week for 10 weeks, you got to start from scratch that next Monday. And that's really,

14:29 really tough. So this more retainer like relationship that I have with my clients now,

14:35 I never want to be working with fewer than three to five clients at a time,

14:39 because it's not like all five of them are going to suddenly reach project end simultaneously.

14:45 All right. I need you.

14:46 So that's the trip.

14:47 Yeah. I've heard of this term called productized consulting, where it's a little bit like what

14:53 you're talking about. You know, maybe if your job is to maintain Django websites,

14:58 instead of saying you can pay me $150 an hour to maintain your site, you could say you can subscribe

15:04 to my sort of site maintenance service and you pay me $2,000 a month and I will jump on a problem

15:11 for up to this amount of time, like almost like a mechanic or a lawyer who needs to jump in and

15:16 solve a problem. And it sounds a little bit like that.

15:19 It's very much like, I think it's very much like a lawyer in that if you adopt that kind of a style,

15:26 now, of course, at this stage of my career that I'm doing somewhat more analytics management

15:30 consulting, right? I'm working with the VP of analytics or some director level. So my particular

15:36 work at this moment in time is a little bit different than what you were describing, but a

15:39 subscription like that is a fabulous, fabulous idea because that's exactly how you want to do it.

15:44 And clients are not going to know if at any given moment you end up with a 65 hour week.

15:51 If a gig comes in and you say, you know, I've got something ramping up and I've got something

15:55 coming to an end. I can power through this one week. I'm going to go ahead and do a 65 hour week.

16:01 And I'm going to do that knowing that if you didn't do that, you might have a 20 hour week.

16:05 Otherwise, right.

16:07 If you're subscribing a subscription model or a retainer model like that,

16:11 you can do that and you just power through the busy times.

16:14 Right. And then you've got that predictable income. You can also do that to have a different

16:19 type of lifestyle, not just to sort of smooth out the income differences, right? You can decide

16:25 I'm going to power. It looks like there's a lot of work these two months. I'm going to just power

16:29 through that. Then I'm going to take a month off and go to Hawaii and I can do that because that's

16:33 just what I want to do. And it's Christmas time anyway. Like work is down. You know, it gives you

16:38 that ability to, to make different trade-offs. Whereas if your job is to be in a cubicle,

16:43 because they want you there for eight hours a day, 40 hours a week, like those trade-offs

16:47 are much harder to make.

16:48 This portion of Talk Python To Me is sponsored by Linode. Simplify your infrastructure and cut

16:54 your cloud bills in half with Linode's Linux virtual machines. Develop, deploy, and scale

16:58 your modern applications faster and easier. Whether you're developing a personal project

17:03 or managing large workloads, you deserve simple, affordable, and accessible cloud computing solutions.

17:08 As listeners of Talk Python To Me, you'll get a $100 free credit. You can find all the details

17:14 at talkpython.fm/linode. Linode has data centers around the world with the same simple and consistent

17:20 pricing, regardless of location. Just choose the data center that's nearest to your users. You'll also

17:26 receive 24/7/365 human support with no tiers or handoffs, regardless of your plan size. You can choose

17:33 shared and dedicated compute instances, or you can use your $100 in credit on S3 compatible object storage,

17:39 managed Kubernetes clusters, and more. If it runs on Linux, it runs on Linode. Visit talkpython.fm/linode

17:46 or click the link in your show notes, then click that create free account button to get started.

17:52 Absolutely. Absolutely. So, and of course, you have the option of working from Hawaii. So,

18:00 what I did for years, now of course, we're all still trying to figure out what long-term, if not

18:07 permanent impact, COVID-19 is going to have on our work lives. I think we all think that we're going to

18:12 be working remotely a lot more than we ever did, right? But we're still sorting out all of that. But for

18:17 years, I would look at a map and look at my calendar and say, "Wow, I've got to be in Munich the third week

18:26 of June, and I've got to be in Amsterdam the first week of June. I'm going to spend the whole month of

18:33 June in Europe." Yeah.

18:34 "Italy. Why not?" "It's not that hard to take daytime U.S. calls in the early evening and just plan your day around that."

18:45 Yeah, absolutely. I lived for a year in Germany and I just had, my work day was like two in the

18:50 afternoon until 10 PM. And it was great. I just had the mornings to myself.

18:53 Yeah. It can be fantastic.

18:55 For the family.

18:55 Yeah. So I think that's the stage for some really nice ideas, some really nice lifestyles and freedom

19:02 and whatnot. But often I think people see it as a switch, right? I hear people talking like,

19:10 "I'm going to save up a bunch of money. I've got this idea. And when I'm ready, I've got enough runway

19:15 that I can go for this long and I'm going to quit my job and I'm going to start this thing." And to me,

19:19 that seems really, really risky. A much better idea is if it's possible, can you build up

19:27 some thing and show that you have traction in it? Because if you can get it to work and you can get

19:33 traction in training and speaking in all these different areas that we might be talking about

19:39 coming up, if you can find traction in there when you're doing it two hours a day, well,

19:43 all of a sudden, if you can put all your energy into it, it's very likely to be successful. But if

19:47 it's not, if it's just going to spin and get no attention, then it's going to really be quite

19:52 devastating if you blow your whole life savings on it.

19:54 Oh, the flipping the switch. I don't think it's too blunt to just say, "Terrible idea." If you can

20:00 avoid it.

20:00 Yeah.

20:01 I'm sure somebody out there has managed to pull it off, but maybe I can help kind of describe the

20:06 context of when, what starting out was like. So I can't imagine that the technical training is ever

20:15 going to go away. Now it's very different now than it was when I started because we have Udemy and all

20:21 kinds of platforms where there's a lot of content. So the kind of classroom training that I did where I

20:26 was teaching classes that were three, four or five days long. You don't see that as much before. But

20:31 when I started out doing that, I was getting paid by the day. And this is something that probably

20:36 people don't think about, but was huge in my life at the time was that since I was in my late twenties,

20:44 teaching these introductory stats software classes, most of my colleagues were full-time, but I was getting

20:50 paid by the day. So I would basically say that I'll teach this stuff anywhere. I don't care if you need

20:58 somebody in Seattle or Texas or overseas or whatever, I'll do it. Okay. A lot of the folks that were on

21:06 salary didn't necessarily have to travel because a lot of times they were assigned to the Chicago office

21:11 or the New York office. Long story short, I actually had more training days per year than most of my

21:19 salaried colleagues had. Wow. So I had a crazy schedule, had an intense schedule, but you can imagine it was

21:25 good money because actually, because I was getting paid a contract rate for more days. And because I would

21:30 sometimes do 13 five day weeks in a row, which is usual for a contract situation like that. What it did is

21:37 it put me in front of a lot of people. And it meant that if some kind of short-term engagement came along, it was

21:43 just a matter of me blocking out two weeks on my calendar. It was really as simple as that. I didn't even have to have a

21:48 conversation. I just had to say first half of April, I'm tied up. Don't assign me anything.

21:52 Yeah. That's fantastic. So let's start there in terms of some of the side hustle potentials. I think

21:59 this professional training, this in-person training, even sans COVID, right? We got to put that caveat

22:05 out there, but yeah, in the normal world, this in-person training and maybe virtual resume or online

22:12 training, this is a pretty good way to do it because for most jobs, I feel like you can make this

22:17 something that you can do without your company completely freaking out, right? So if you're a

22:23 data scientist for say a tractor manufacturer in the Midwest, you could make the case like, let me

22:28 take an unpaid week every two months and do one of these courses. It will make sure that I'm on top of

22:35 my game and that I'm a better person for you. And I can help give some internal presentations as part of

22:40 my regular job and stuff. Like that's a pretty good story. You can tell without flipping the switch

22:45 that will get you exposure to these clients that you can reach out to if you want to go further and so

22:49 on. What do you think about that?

22:50 Absolutely. So let me amplify that even more. I think if someone were to have this kind of

22:56 conversation with their boss, this isn't a crazy scenario at all. Hey, there's this cool conference

23:01 happening out in Vegas. There's a chance that I could present one of our client case studies and kind

23:07 of promote the company. They're not paying me to give the talk, but they, you know, I've talked to the

23:12 conference organizers and they think my proposal would have a good chance of being accepted.

23:15 I get to talk for like an hour and then don't just do that. Right. You might then have free travel

23:21 and you're getting paid. You're on your salary, given that presentation, you know, for the boss,

23:25 but then take maybe just like two days off, but stick around and hang out the rest of the conference

23:31 because a lot of travel budgets are a lot of times they sell, you're going to give a one hour talk on a

23:35 Monday, then fly out on the Sunday and then, you know, fly back on the Monday night, which very early

23:40 on, I just knew that was not my style. Right. So yeah. Stay an extra day, sleep in and just have a

23:47 day where you just travel and just chill and don't make it a stressful experience.

23:50 Well, I'd want to get professional development in there too, though, which is like a key point.

23:54 So sure. Do some Vegas. Everybody thinks of a casinos, but there's also some awesome hiking

23:59 out there. You can really do the outdoorsy thing in Vegas too, if you care to, but I would definitely

24:03 want to make sure that I was attending the conference too, not just speaking and then

24:07 rushing back. But then on top of all of that, see if there isn't a way to score a pre or post

24:14 conference workshop. Most conferences have them. Okay. So maybe somebody is really fantastic at

24:21 particular Python package, or they've got some skill. Maybe they're an early adopter of something new.

24:27 Reach out to the conference organizers and say, I noticed that you've got a post conference

24:32 workshop and X, Y, and Z. Would you add one on my topic?

24:35 Yeah. Or even be a little more preemptive about it and just like apply to do that workshop far in

24:41 the future. Right. Right. When the call for proposals or call for papers opens.

24:45 Sure. You absolutely could do that. But in my experience, sometimes the paid stuff is curated.

24:51 Okay.

24:52 So it's not so much proposal based. It's more like the draw, right? All conference organizers need

24:58 two kinds of content. They need the kind of content that they offer so that everybody gets a chance to

25:04 speak. That's like the case studies and stuff. And then you need the draw. The keynotes are being paid.

25:10 And that's presumably why people want to go to the conferences to see the keynote. With concurrent

25:14 sessions, let's face it, for all to be honest, we kind of cross our fingers with one of those concurrent

25:19 sessions and we hope they're good. But usually we don't expect that much. We just hope that we're

25:23 wrong about that. It turns out to be great. But you expect the keynotes to be great and you expect the

25:27 pre and post workshops to be great. So in my experience, you have to kind of sell yourself

25:32 on the conference organizer that one, they want to add that topic and that you'd be a good person to

25:36 present it. But if you do that, now you're getting paid to go to the conference by your boss.

25:41 You're taking a couple of days off. And I would be upfront with my boss about that kind of thing.

25:46 Because they were like, somehow you talk them into giving a workshop on that topic.

25:49 That's fabulous.

25:50 You know?

25:51 Yeah. Yeah. And I think the story you tell is this is making me a better developer, better data

25:57 scientist for as one of your employees, I'm coming back better. Right. And it doesn't cost you

26:02 anything. And yet it builds these relationships and it builds these funnels. So let's talk just a

26:08 little bit more about getting in the training side of things. So one of the challenges is I want to

26:14 get into training is making those connections. So you can do speaking and you can do blogging and

26:18 other stuff to make those connections. But then you get a big company, they call you up and say,

26:24 we want this. And that's interesting, but we also need this variation and that other thing.

26:28 And you have very little materials to present from, right? You've got to go and basically write

26:33 your educational materials. Or you can go and work for some established training company that

26:40 already has those relationships, already has the materials and just need somebody to put it all

26:45 together, someone to actually do the presentation. So that's another option. There's so many companies

26:50 that have these like very loose consulting relationships with experts, right? Places like

26:57 Wentelect and other companies that, right, they've got the sales team, they have the relationships,

27:01 they have the materials generally, but they need people to, who are experts.

27:06 Yeah, that's all very true. But I think the first step, the first step isn't the entrepreneurial

27:11 step. That's more like the second step. The first step is you kind of have to be almost like a secret

27:17 shopper and research in a topic that you feel well-versed in or that you're super intrigued with,

27:23 right? I mean, it could be something you haven't mastered yet. You're, you got some confidence,

27:27 but you're super excited about it. You have to research how people that want to learn more about that

27:32 topic, get trained because that's indirectly going to lead you to what that training market is like.

27:38 So it could be the vendor that, for instance, I use RStudio. I know they do webinars and things.

27:46 I don't, I've never researched that market, like how much RStudio does workshops. I'm sure they do,

27:52 but that's where you would start. You'd be, how many conferences are there a year? I'd even get very

27:59 specific about it. I would say, oh, I found this cool topic. How many times can I find this offered

28:05 a year? Is it always the same instructor? That's one of the first things I look for. Is it somebody

28:10 that is in one particular topic? Let's say, for instance, I happen to be a GG plot two fan. So

28:16 obviously sometimes you'd have Hadley Wickham given some kind of GG plot two talk. If he's the only one

28:21 that's talking about it at all the R conferences at this point, there'd be enough demand for it that that

28:26 would no longer be the case. But if it's always him and you're going to be competing with Hadley

28:30 Wickham teaching GG plot two, not going to happen. Right. But I doubt that's what you would find if

28:35 you research that in 2020. I don't think he'd be the only one teaching GG plot two. So you'd seek out and

28:40 say, okay, how many conferences offered something like that? Were they part of the conference fee?

28:45 Were they a paid event in addition to the conference fee before, after? This is all the kind of stuff that you

28:51 have to know because naturally if it's part of the conference fee, it's probably free. Well,

28:57 it is free to conference goers. So therefore the conference organizer doesn't have a lot of cash to

29:02 spend on that. If it's a special additional expense before, after, and it's a topic that you're really

29:09 good at and think you could present well, what may be happening is sometimes they can't find someone in

29:15 a particular city to present that. So that's another thing too, is if someone lives in a medium

29:19 to large city, look for events coming to you. Maybe there's expert that isn't available to travel to

29:26 that one. If you're local, you don't have to charge travel to the organizer and they're going to like

29:31 that. Yeah, they do like that. That's definitely something you could promote to at least conferences

29:35 that cover travel and say, look, you don't have to pay for me. I'm already here. Or I can get myself

29:40 there with a car or a train for almost nothing. Another thing in this realm, while we're on this

29:45 side of the things is webcasts, webinar type things. So that's a lower bar, I think, often than speaking

29:54 at a conference, right? You can approach various groups, right? Like I know JetBrains has a bunch of

29:59 external people who give presentations on stuff. And if you're really good at some aspect of data science

30:05 and you're willing to like use their tools to present it, often they'll say, yeah, sure. We'll

30:10 use our platform and our audience to let you do that presentation. We just did one with Chris Moffitt

30:15 and had 1,100 people sign up with sending out one email and one, a couple of Twitter messages.

30:21 That's a pretty big group of people at all. We didn't have to organize a conference or convince

30:26 anyone. We just put it on the internet and told people about it, right? So that's also an option.

30:31 I agree. Now, the difference there is that there's a good chance that something like that

30:38 isn't going to get paid. I mean, obviously, sometimes I've been paid quite a few times to do

30:43 a webinar, but I've also done a ton of free ones, but absolutely to do it in terms of the visibility.

30:49 And you've got to think that that option is just going to increase in frequency because whole

30:54 conferences have gone online with COVID-19. And I think that may be permanent in that some conferences

31:00 will choose. I have this crazy theory that some conferences will go every other year,

31:05 every other year live and every other year online. I've never heard of anybody saying that,

31:09 but I just have a feeling that's going to happen. The other thing too, is that there's usually a team

31:14 for vendors that's in charge of such things. Like KNIME has what they call the evangelism team.

31:20 Those are the folks, they're employees of KNIME that write the training materials and do various

31:25 things. But think about it. If you belong to that team, you don't always want to be the face of that

31:32 like all the time. You know that if you mix it up, it's going to be of interest. So if someone's willing

31:35 to volunteer an hour, now granted that hour webinar is going to take you a day or two to prepare.

31:42 It may be stuff that you already know, not that you don't know it, you're going to prepare examples

31:46 and so on. So it's going to take some time, particularly if you're new at that kind of thing.

31:50 But if you just simply found who the training team or the evangelism team with an organization

31:55 like KNIME and just said, Hey, I've got this interesting topic. Do you have any webinar slots

32:00 coming up? I'd volunteer my time. I just, I'm excited about the topic and I want to share it

32:05 with the world. They're not going to turn you down. Yeah. Usually they're just looking for content like

32:10 that. Like, what are we going to do this month? I don't know. It's another month we're starting over.

32:14 Awesome. All right. So I definitely agree that those kinds of things are awesome for the visibility,

32:19 not for direct income, but they build that foundation that you can sort of branch out to.

32:23 So we've covered training. We've covered conferences, consulting, I think mixes in there. We talked

32:29 about that. You mentioned books, but what about writing? Give us the story where you think the

32:34 data science side hustle of like writing books, maybe even magazine articles, if those are still

32:40 the thing. I know there's some that do pay actually for that. Even maybe reviewing a book.

32:43 Yeah, no, that's interesting. Yeah. I think of like freelance writing like magazine articles. I think

32:49 of kind of people that spend 10 years sending short stories to the New Yorker and then I don't think

32:54 everyone picked up obviously in the tech now where it's very different. Exactly.

32:58 Forbes. I haven't written Forbes, but they have an interesting model or even medium, things like

33:04 that. So that's not an area where I have a lot of experience, but absolutely. I think that is

33:08 available. I think I should probably briefly tell the story of my first book. So it came out in 2013.

33:14 When would I would have started on that? The first one tends to take forever because you're learning

33:19 the ropes of just how books work. So I think that I probably wrote the proposal for that in 2011.

33:25 So I would have been about 12 or more years into my career from like the busy part of my career even.

33:33 And I was just always kind of scared to do it. I didn't even know how to start. But since I was well

33:39 known in the SPSS model or community, I reached out to, I don't know if everybody has heard the

33:44 acronym CRISP-DM. It's the cross industry standard process for data mining. It's very much on the

33:50 machine learning side. So some people may be familiar with it. Okay. I reached out to one of

33:54 the coauthors of CRISP-DM because I knew him. He knew me. And I said, why do you think it is that no

33:59 one's ever written a book on the basics of using SPSS modeler? And I knew him. I wasn't afraid to reach

34:06 out to him, but I did think that the chances that he would want to collaborate on something were slim

34:11 to none because he was so much more well-known than me. He was the author. It would be like,

34:16 it'd be like doing something with a coauthor of Agile or something. You know what I mean? Just by

34:20 definition, he was well-known within the community. And what he said just stuck with me because I think

34:25 it's so often true. He said, I kind of always wanted to do it myself, but I was just too busy.

34:31 So if you're less experienced, less well-known, but have the bandwidth, reach out to these people

34:37 that are expert and well-known because usually they have those two things, but they don't have

34:43 the bandwidth. And then there you go. You could pair up. So we put together an outline. I was all ready

34:49 to shop it around to publishers, something that I didn't know the first thing about how to do. I truly

34:54 didn't. I was completely lost in this project. All I knew was that I wanted to do it. And I had such an

34:58 amazing coauthor lined up that I felt like I couldn't fail. And out of the blue, I got a

35:04 LinkedIn message from a publisher that wanted to do a slightly different book. Now, keep in mind that

35:11 no one in the world, except for me and this one other person knew that we were working on an outline,

35:16 but it's just so funny that at that point in my life that I felt I was ready, somebody else decided I

35:20 was ready. So we skipped the intro book or rather we delayed it. And we ended up doing my first book,

35:27 which is a IBM SPSS model or cookbook. So the whole idea of a cookbook, and of course,

35:31 all the coders will know this is lots of short examples where you address something very specific.

35:36 Right.

35:36 And it's not like it's sold like Harry Potter, but since it was the first of its kind and the

35:43 publisher knew that, that's why they were looking for authors. See, that's what people don't realize

35:47 is that publishers have topics that they want addressed and they are seeking out authors to do that.

35:54 I think it's just like when you apply to college.

35:55 I get like every other week I get contacted by somebody. Hey, Michael, I'm the acquisition editor for so-and-so.

36:02 We're looking for a topic, a book on this. Are you interested? No, I'm trying to run my business.

36:07 Yeah.

36:08 You're going to have to find someone else. I'm sorry, but this is just not. But yeah,

36:11 you're right. Because I still see that. I saw that last week.

36:14 Yeah. Yeah, absolutely. And the thing is, well, you and I think have probably figured this out at this

36:20 stage in our careers is that the reason that they're reaching out to you is the things like this podcast

36:26 that you're doing. Right. But now that you're doing those things, you're busier. And not only that,

36:33 the other thing that people don't realize is that they probably think the first time they get an email

36:37 like that, oh my gosh, am I worthy of this inquiry from a publisher? They really want me to write a

36:45 book. That's so amazing. And then you realize eventually that it's almost like a spam thing,

36:49 that the concepts and publishers and book topics vary in quality. And then eventually you find out

36:55 that not all of them are worth doing. But we put it together. And it was with, frankly,

37:00 one of these publishers that does reach out in that way. It's Pact is the name of the publisher,

37:04 because their whole strategy is they're super aggressive about the search engine optimization

37:08 thing. If they see that something's coming up in Google search and there's no book written about it,

37:13 they want a book about that right away. They want to be first to market.

37:16 They're on it. Yeah.

37:17 So it's not like getting a contract with Random House or Wiley, or when you hear about these

37:25 politicians or famous people getting million dollar book advances, that is not the kind of book

37:29 deal that we're talking about, right? But they know that there's a hole in the market and they

37:35 want to find somebody to do it. So as long as somebody has a reasonable amount of competence,

37:39 that's going to be enough to be able to score the book deal. And what we were talking about earlier

37:44 is sufficient. I did three webinars on the topic. I did one conference workshop.

37:49 Yeah.

37:49 Somebody could achieve that in a year or two, and that would be sufficient. So I could have written

37:54 my first book years before I did it, but I was afraid to try.

37:58 Right. And if you don't have an audience, it's not a bad route to go. I think one of the really

38:03 big challenges of so much of this is getting awareness and getting attention and getting people

38:08 like you might build something amazing or you might create an amazing book or course or something.

38:13 But if people don't pay attention to it, it doesn't matter. Right. It's just, it's such a busy

38:18 world. So like for me, it doesn't make sense for me to try to partner up with one of those guys and

38:24 spend a year and a half writing a book because I can reach so many thousands of people weekly through

38:29 the various podcasts and other formats that I have. But if I was working at a company, passionate about

38:34 a thing, but had basically zero following, it seems like a good choice for at least the first time.

38:40 Well said, really, because you're not going to make money. You're not going to make money from

38:43 the book. Not these kinds of books, right? The only kinds of books where I'll meet authors that

38:49 make a pretty good income from the book is more the Harvard Business Review type on leadership,

38:56 communication. I guess one of the bestselling books of all time was The One Minute Manager.

39:01 Somebody was just telling me that a couple of days ago. That's not the kind of books that as data

39:05 scientists we're going to write. So your mileage varies as they say, but we're talking,

39:11 you should be able to make a few thousand dollars, but, oh, I've got a great way to actually kind of

39:17 bracket what can be expected. Years ago, don't know her last name, but the Linda of Linda.com. I'm

39:24 embarrassed that I can't remember her last name because she's famous. She started a training business

39:28 and she was one of the first to realize that, oh, maybe we have to do like videotapes and DVDs.

39:33 This was not that long ago, but she started-

39:37 Yeah, this was 15 years ago, maybe. Yeah.

39:39 Maybe a little bit more than that. Could be-

39:42 Could be 20.

39:43 Yeah, could be 20. And of course, she ended up selling to LinkedIn for 1.5 billion,

39:48 which is a really nice paycheck at the end. Not everybody's going to get that.

39:51 But the reason I bring her up is because if people want to get an idea of what a home run

39:56 is in the technical book space, because I don't know what people's expectations are, right?

40:01 But years ago, when Linda and her husband wrote their first technical book, I think they made

40:06 something like 60 or 80,000. So you're taking into account inflation, maybe that'd be like 100,000.

40:12 That would be considered absolutely rock star, amazing, amazing, amazing royalties for a technical

40:19 book. Most people are going to be a 10th of that or less. Yeah.

40:23 And I've read that a lot of books don't even sell 1,000 copies the first year. So that's an absolute

40:30 home run. So someone that thinks that they're going to retire on a book or something, not going to happen

40:35 because that might be one of the most successful technical books in recent memory because it's

40:41 notable because she used that to start her company. Don't count on that happening.

40:44 Yeah, absolutely. I think so many of these things can be thought of as layers that you add on that

40:51 allow the next step to be taken, allow the next thing to be successful, right? You're not necessarily

40:56 going to get rich off of that book, but if you write the book and then you get well-known, you do some

41:01 speaking, and then you do this other thing, you do the startup, well, you might be in a much better

41:05 place to have that startup be successful if you hadn't done those things.

41:08 Absolutely. You know, I had a conversation. It was a brief one-on-one conversation.

41:13 It was a private conversation. It wasn't in a public setting, but I don't think there's any

41:17 harm in sharing it. I was chatting with Michael Berry, who some people may know because he kind

41:23 of does the conference circuit from time to time, but he's now head of business to business on TripAdvisor.

41:29 So he's a very successful data scientist. His co-author and friend, Gordon Lenoff,

41:35 has got a senior data science position at the New York Times. So they're both like super well thought of.

41:39 He was in one of my training classes. Now, of course, at the time, he was already an author and much more established than me. But the reason he was sitting in was

41:47 purely because he needed to learn the tool. So he was at the time a SAS person and wanted to learn

41:53 the SPSS equivalent. So you can imagine having a well-known person like that in class at the end

41:59 of each chapter. I would say, well, Michael, would you like to Michael's comment on Keith's chapter,

42:04 you know, kind of a thing. It was kind of fun, but he approached me during a break and he said,

42:08 Keith, you know, you seem to be pretty good at this. I noticed that your training,

42:11 I'm kind of surprised that you're not consulting. And the answer that I gave him was the thing about

42:15 consulting that's working for me right now is I'm doing a five-day week after a five-day week after

42:21 a five-day week. I mean, I just keep on getting my daily rate. And there was an age difference,

42:25 an experience difference. So I felt just a little bit bold to say, and I'm guessing that the way you

42:30 and Gordon have structured your consultancy is that you probably aren't billing 22 days a month.

42:35 It probably doesn't work that way. He said, you're right. He said, but you know what happens is when you have gaps in time, you have to use that effectively.

42:43 And what he and Gordon did is wrote their first edition of their book. They actually wrote several

42:48 books, but they said that was just when, whenever there was a week and the phone wasn't ringing,

42:51 we would chip away, chip away, chip away at the book. And when that first book came out,

42:56 he said, the phone never stopped ringing after. We were slammed after the book came out.

43:01 So, and you know, that was their first book would have come out almost 30 years ago now.

43:06 And they've been full-time data scientists, freelance and salary that whole time. So it

43:12 really was the beginning of, they were already well-established, but the beginning of a very

43:16 successful career was writing that first book. So that's why you do it.

43:20 Yeah. Very interesting. Okay. So let's focus on a couple of other areas that I think

43:24 will be very exciting for people. I guess there's always the startup, right? And I think there's

43:31 some interesting areas in which data scientists can create startups, right? You can go and pull

43:36 different data together using screen scraping, using APIs, you know, all sorts of stuff. And you can

43:41 come up with inferences and trends and expectations. You can sell that data back to people, right? I think

43:48 that that's a really decent way to maybe create some kind of API or some kind of system where you really

43:55 want to know about this market, not just this one thing. We've brought it all together. We've got the

44:00 infrastructure to keep it up to date and it's worth you to pay a hundred or a thousand dollars a month so that

44:05 you just get that data and you don't have a team that has to track it down and clean it and deal with it.

44:10 Yeah. And you know, this is really interesting because if you think about the story that I just

44:15 told about Gordon and Lenoff and in my own experiences being in my late twenties and the late

44:21 nineties, this obviously wasn't an option that was available to me at the time, but I can see that

44:27 absolutely this route could be the equivalent of that first book for many, where if the app is successful

44:35 enough that it's a worthy component of their portfolio, they don't have to retire on the app.

44:42 The app could be the basis on which they become a freelancer. So again, it wouldn't be something that

44:48 was available to me when I was starting out as an option, but I think it could be the equivalent of

44:53 that. I mean, if someone is more excited about a startup option than they are about writing a

44:58 technical book, I mean, if the notion of reading a technical book leaves them cold, and I'm sure for

45:02 some people it does, you know, if this is what they're passionate about, it can absolutely be the

45:08 same credibility builder that a book would be because that's what you need. I mean, there's no

45:13 question that the only reason I get paid to speak, whether it be a keynote or a conference workshop

45:19 with regularity, it's not enough to be my whole year, but I've never pursued it to be enough. But

45:24 on the days that I'm doing that, I'm making good money those weeks that I'm focused on that.

45:29 There is no question that the reason I can do that is the books. So everybody needs a credibility

45:35 builder and apps and startups could be that.

45:38 For sure. Let me lay out another path. I think this, there's a lot of cool ML stuff and predictive

45:43 stuff that I think that if you could bring enough data together and you could predict the outcome in

45:48 some market or in some area better than anyone else, you can sell it. I'm pretty sure that you could find

45:54 a way to sell that to companies that want that prediction to be made. And they're already asking their

45:59 people to do it, but they're not doing a good enough job. And if you got specialty in an area,

46:03 chances are that area doesn't have this thing. That said, another interesting path I see,

46:08 especially in the Python space is popular, well-accepted open source library creates a

46:16 open source plus company. So examples of this are the Scrapey API for web scraping. Those guys went and

46:24 founded Scraping Hub, which is web scraping as a platform service, right? So they've got the

46:30 infrastructure and they've got the distributed, whatever, got all the databases and the caching.

46:35 And you can just say, I want to take that library that you wrote. It's open source, but I want to

46:39 have you run it because of all the challenges of getting blocked and whatnot. What's scraping?

46:43 Matthew Rocklin started along with Hugo Brown Anderson. They started Coiled, which is

46:51 Dask distributed Python computing as a service. So Dask is open source, free library.

46:56 Coiled is the Dask as a service. We have Explosion AI, which is spaCy, the ML open source library,

47:04 but Explosion AI is like more tooling and labeling stuff on top of it as a paid service. So over and

47:11 over again, I'm seeing these companies that are founded by the lead person of some really popular

47:18 open source library in a way that doesn't undermine the open source aspect of it, but builds on it and

47:24 says, you really love this. Well, here's what you can do with it. Now that's a pretty long runway

47:28 because you have to have a popular open source library, but having a popular open source library

47:32 is also something you can sell to your company and do as a side hustle pretty well. Because very rarely

47:37 are like, are you creating a library that does web scraping? We're going to have to fire you because

47:42 you're doing that in your spare time and it doesn't make sense. Like this is such a conflict of

47:45 interest. You just don't hear that very often. I mean, certain circumstances maybe, but it seems like

47:50 something you could do pretty easily. And if it's really successful, then you're ready to take that

47:53 step on top of that. I agree. Being a somewhat old school machine warning guy, one of the topics that

47:59 I'm incredibly passionate about is feature engineering. How can I add that little bit of spice to a model to

48:07 get it to take off? And as I was listening to you, I was reminded of the fairly recent Zillow competition.

48:14 And I guess there were three guys that end up ensembling their solutions. But what I remembered

48:19 was that one of them just concentrated on one component of the problem of the home value,

48:24 this one piece. And that was commuting distance, not distance like we're all flying to work in a

48:31 helicopter because most of us are not, you know, but dealing with public transportation and all those

48:36 variables. So he would take all these addresses and just try to crack the code on that. No one else

48:41 on his team did that. So I can totally see how if someone just focused upon addressing one small

48:48 piece of the puzzle, building upon available open source tools, that one, again, that's credibility

48:55 building, but also I agree with you that it could also be marketable.

48:58 Yeah. Yeah, for sure. That's an interesting challenge because it's not distance. It's not

49:02 even following the map distance. It's like time. Time is the thing you actually care about and commute,

49:07 not actual distance, which is not so easy. Exactly. That's what impressed me too. Yeah.

49:12 Yeah. As somebody who has been working from home for a long time, I'm very happy. I don't have to

49:17 think about that. All right. Let's talk about some other areas. You mentioned the Zillow thing.

49:22 There are some competitions for data scientists that I think have monetary payoffs, right? I think Kaggle

49:28 has some monetary payoffs and some other ones as well. What about that?

49:31 Yeah. Well, again, there's always, I think if you shoot for the credibility and you get the

49:36 credibility and the money, it's kind of like one leads to the other. So Kaggle is an interesting case

49:42 because there's the big prizes, but H2O.ai, open source company that a number of folks are probably

49:49 familiar with, they have an advisory board that they've assembled of a bunch of Kaggle grandmasters.

49:55 And it's really something because you go to the conference and there's about 15 of these folks up on

49:59 stage. But the other way that they're using them isn't just to parade them around during the

50:04 conference, which is for me, I'll apologize to them. It's slightly humorous to have this.

50:08 It's almost feels like, you know, they all like march out and you got 15 people on a panel. It's

50:12 somewhat of a little bit crazy because how can you have 15 people on a panel? But the other way that

50:17 they utilize them is they ask these folks, what would you do in the context of their auto ML product

50:24 development, trying to imitate using them almost like you would develop an expert system

50:29 to build models. So what's occurred to me is that you can come in and the top hundred,

50:35 even probably in the top thousand and some of these Kaggle competition, and you have bragging rights,

50:41 even if you don't win the big prize. And even if you don't win the big prize, there could be,

50:46 because I'm sure these advisory boards are paid. I have no doubt if they're flying around

50:50 to appear at a conference and stuff. They're getting paid to be on that. So you might be

50:54 reached out to for advice. In other words, that's what paid advice. That's all that what consulting is

51:00 because you did well in a competition, even if you didn't win.

51:03 Yeah. Very interesting. I agree. One of the areas I feel like making sure winning scoring high in a

51:09 Kaggle competition is quite challenging. Creating an open source project and then founding the

51:14 successful open source commercial business on top of that is quite a big step in a long runway.

51:20 Something that's much shorter that people could get started with sooner would be mentoring.

51:25 I feel like mentoring is a low-grade consulting, right? You don't have to come in and say,

51:31 do your Fortune 500 company? I have this expertise better than your employees and all that. This way,

51:36 usually I get messages like this all the time. Hey, Michael, I'm really trying to do this thing.

51:40 I'm a little bit stuck. I've been here for a while. It's frustrating. Can you help me? And I almost

51:45 always have to say, no, I can't because I'm so busy doing all my other stuff. But there are places

51:51 you can go where you can spend an afternoon helping somebody get paid for that and sort of develop that

51:58 experience of more corporate consulting type of model. Yeah, definitely. And there are a lot of folks,

52:07 including quite a few in their 20s that have done YouTube channels or various things on the side.

52:15 And I increasingly see that you'll encounter somebody like that and they'll have these calendar

52:24 links. There's tons of apps where you can do a calendar. And they'll basically just say,

52:29 if you have a question, pick a time. Yeah.

52:31 A hundred bucks an hour on up kind of a thing. And that's definitely an interesting way to do it.

52:36 So I would think that if somebody was starting with zero reputation and they used one of these

52:43 sites, Upwork or what have you, and people complain about the rates and that it's not like a great deal.

52:48 But if you did that for six months and tolerated the fact that you were getting paid,

52:54 maybe not great, but getting better at it, I don't think it's a huge leap to go from that

52:59 to part-time Python help desk in the evenings or in the weekends for another six months or a year.

53:07 Now you're making a good hourly. You're just not doing a lot of it. And then next thing you know,

53:11 you have a mailing list. Next thing you know, you have LinkedIn followers. And there are folks that

53:16 do that kind of thing full-time.

53:17 Yeah, absolutely. The YouTube thing is interesting. Let's come back to that.

53:20 For mentoring, I have not used it personally, but codementor.io seems like a pretty good

53:26 two-sided marketplace. You go there, say, I have expertise in this and I'm willing to other people

53:30 say, I need help in that. And it'll match you up. Hired and places like that are also pretty good at

53:36 making those connections for consulting setups. Upwork's interesting. So I hire some people through

53:42 Upwork and I've had really good experience for the most part. People I worked with,

53:47 for years through Upwork and it's been fantastic. I think the software development side of Upwork

53:52 is a bit of a mess. I'm not entirely sure. But a lot of times people ask me, how do you get started?

53:59 There's this chicken and egg. I'm only going to hire somebody who has experience, but if I'm new,

54:04 how do I get experience so I can get that job that requires experience? How do you make that?

54:08 How do you break that cycle? And to me, I feel like those places like Upwork actually have a really

54:12 good opportunity for you. Like you over there and if you get paid $15 an hour to build something,

54:17 that's going to sound like it sucks. But if in one month you can say, my experience is I built that

54:23 thing over there and I can show that as part of my resume, all of a sudden you're no longer a

54:28 completely unknown bootcamp graduating data scientist. You're now a professional developer

54:34 with stuff published in production you can talk about. Right?

54:37 So I almost see that as an opportunity to get one of these, like an internship almost. You just got to

54:44 kind of, you know, it might not be the best experience and it might be it's not the right

54:47 way. But for people who are having a hard time, I feel like you can use places like Upwork to

54:51 like take a half step into professional development and then use that to take the next step and get going.

54:57 I agree. I think you just have to, if I was, because again, my first related gig was teaching the

55:04 software classes. So that's how I got my foot in the door. But if I was looking for a way to

55:09 establish some credibility, I mean, getting paid something to build up a portfolio seems to me

55:15 better than unpaid unless the unpaid is a fabulous opportunity, which some intern unpaid internships

55:21 might be right. Yeah. So I think you just tolerate the fact that the pay is not great for six months

55:27 because then you have a portfolio, but be looking for the on ramp for something better. But I don't

55:31 think it's crazy. I mean, there's a lot of negative feedback about it because, you know,

55:34 the rates aren't great or they take too much. A lot of the feedback sounds not unlike what Lyft or Uber

55:41 drivers would say is that the reality is once you take into account fees and deductions and everything

55:47 that you'll be disappointed with what you're left with. But I agree with your characterization,

55:51 which is you're building a portfolio. It should be quite short term. I can't imagine that somebody

55:57 would want to do that for more than six months to a year.

56:00 But I do think it allows people to break that cycle of I have no clients, I have no experience,

56:05 and I'm looking for that job where they say you need to have some experience in this. Like there's

56:08 a not just a thing I did on GitHub as my free time, but there's a professional thing in production I did.

56:13 You don't have to say how you got the job. You don't have to say how much you were paid for it.

56:17 You just say, I worked as a professional data scientist for six months on that. Right. And okay,

56:21 great. Looks like, you know, let's do it. Right. You're hired. So I think that's fantastic.

56:25 People should not discount that in my opinion. All right. Really, really quickly. Let's riff on

56:30 one final thing. And then I have a, an analogy for folks, I guess. What about teaching at like a

56:35 community college or part-time at a university or something along those lines?

56:40 Well, I've been doing that now for about five years and I find it to be really, really valuable.

56:45 I doubt that. I think this is more a mid-career option because obviously you have to have some

56:52 credibility. Right. But yeah, I find that it just, it seems to be, I mean, I can really only go based

56:59 on like body language or something like that, but it seems like when this comes up in conversation,

57:04 there's a little bit of that. Wow. You do some teaching for them just seems to be kind of like

57:09 a cool thing. So I know there's a lot of demand for it. So you don't have to be the most famous person

57:15 in the world and the topic that you're going to teach, but you do have to be mid-career. So again,

57:21 it's where just one or two paid workshops at a conference, you've done one book and that one book

57:28 maybe sold 800 copies for a not so amazing publisher or whatever. Right. I mean, it doesn't have to be

57:34 amazing, but things like that taken in the aggregate, right. Would be enough to be able to teach

57:39 certainly a basic class or something because there's a lot of demand for it. And then you're

57:44 getting quite a bit of credibility associated with that. And data science is interesting because at least

57:51 for the time being, you don't need a PhD because until recently there weren't PhDs in data science.

58:00 Exactly. So I only have a bachelor's. I don't even have a master's and I teach several courses

58:07 a year for UC Irvine in their certificate program. But again, it's based on my years of experience.

58:13 You know, so I would say how many years would somebody need maybe five or eight or 10, but it

58:17 could be, it could be what separates, you know, in the mid-career phase, it could be what separates you

58:23 from getting accepted to give conference talks from time to time and getting a chance to do your

58:28 first keynote. I don't want to be overly dramatic about it, but that's the kind of thing that might

58:32 make a difference because a conference organizer might want to say, not just Keith's been doing

58:37 this for five years, but he's written X number of books, teaches for UC Irvine, has a LinkedIn

58:42 learning class or whatever. So you never know what that first puzzle piece is going to be falling

58:46 into place. Yeah. But they all build off each other. Yeah. Definitely want to emphasize that

58:51 these things layer and that they're additive a lot of the times. You touched on YouTube just really

58:56 quickly, I guess. That's actually a really interesting thing as well. You could create a YouTube channel

59:00 that basically does live coding of a bunch of stuff or does like a live code review. Let's walk through

59:05 a popular open source libraries code. We'll do another, a different one every week like that. Those kinds

59:11 of things could build credibility and build an audience, which once you have an audience, so many of these

59:16 things become not just possible, but relatively easy. But having that audience is dramatically hard to get.

59:23 True. This is true. So I've been at this for many years. We've established that. But when I was starting

59:29 out, of course, social media wasn't a thing, right? So it's not like I wasn't aware of it, not paying

59:34 attention to it. But it's really when I started doing LinkedIn learning courses that I really started to

59:38 focus on that because my assumption is, I'm sure it's correct, that the more LinkedIn followers that I

59:44 have, the more people will find the courses and vice versa. So I've been, this year in particular,

59:51 since I have not been traveling with COVID-19 going on, I said, well, here it is. I've been

59:56 thrown into a situation as traumatic as COVID has been at times this year that it feels almost like

01:00:00 a sabbatical. This is the first time I bought this house in 94. And this is the first time in all those

01:00:07 years that I've been home for six months straight, I believe. Yeah. Isn't that crazy? Yeah. 85% travel for

01:00:12 20 years. So I said, okay, well, here it is. If we're going to be work from home, at least for the time

01:00:16 being, if not for quite a long time, then I'm going to focus on social media. So I've been learning how

01:00:21 to grow followers and all the various things that you do. But I have in my journey of figuring that out,

01:00:27 I've met folks half my age that have a hundred thousand followers because whereas I was busy doing

01:00:35 other things in the first few years of my career, that's one of the first things that they figured out

01:00:40 that they leveraged. So that in and of itself is a skill. And if someone's going to go that route,

01:00:45 there are ways to crack the code on that. And I've met people that have gotten that many followers in

01:00:51 two years. Yeah. Amazing. Okay. Let me leave you with an analogy that I'm fond of. You give me your

01:00:57 impression on this. So all of these things being side hustles would be typically taken as a side project

01:01:04 while you maintain your regular job that you were then going to either just keep doing for additional

01:01:09 money, leverage and do a better job at some point, or you could quit and make it your full-time job

01:01:13 if it was super successful. Right. But at the moment, the side hustle part is on the side. So extra work,

01:01:19 right? So one of the analogies that I like to think of, because I went through this somewhat with my job,

01:01:24 I was working full-time at least in a pretty intense job, started the podcast, worked on it a bunch,

01:01:31 grew into another podcast. Now I have two podcasts and eventually quit that job and use that foundation

01:01:37 to start my online training course. Now we have authors, we have a bunch of courses and so on.

01:01:41 Right. So, but for the first year of that, that was, let's do these two things at the same time.

01:01:47 And if the other one takes off, maybe I can put more energy into it. Right. I didn't really

01:01:51 necessarily know that I would, but it was fun and let's see where it goes. So my analogy is to do with

01:01:57 rocket launches. So when a rocket takes off, it's, you know, full power as it takes off,

01:02:01 it's going and going and going and it gets going faster and faster. It gets to just about a thousand

01:02:06 kilometers an hour and they have to like power down. I call it max Q where there's the maximum

01:02:11 aerodynamic pressure on the rocket as it goes through the sound barrier. So it's like under the

01:02:17 most pressure. And then eventually once it gets through that barrier, it's easy for it to go.

01:02:20 They just power it back up and it goes up to 18, 20,000 miles an hour or whatever. It just takes

01:02:25 off, right? It goes out into space. So I kind of think of the side hustles in this analogy of max

01:02:30 Q. Like at first it's going to get harder and harder and harder to do both of these things.

01:02:34 And you're going to be pushing against the air or whatever. And it's just like,

01:02:39 ah, this is so frustrating. But at the same time, like if you can make it successful enough that it

01:02:44 could be your thing. And you can, like I said before, if you can do it in two hours,

01:02:47 what if you had 10 hours a day and all of your energy, not when you're tired at the end of the

01:02:52 day from eight to 10 PM, you could, if you're successful with two hours, you can definitely

01:02:57 be successful with more time and energy. So think of it as max Q, right? You've got to like go through

01:03:02 this early stage, like extra hard journey of trying to juggle both things, but eventually it's going to

01:03:08 lead to something better.

01:03:08 I agree. I've never thought of it that way. That completely resonates with me. And I have a couple

01:03:13 of quick examples and response. If you just look at my year this year, okay, my last travel gig,

01:03:21 you know, this COVID-19 year here, 2020. So if my last travel gig was in March, that was chaotic

01:03:27 because of COVID-19, right? I get home and my coauthor, Jesus, and I had to work on the fourth

01:03:33 edition of one of our books. So while that was going on, I wasn't paying attention to other things as much

01:03:40 as I should, because you think fourth edition, aren't you just adding a paragraph here or there?

01:03:44 It's not really, you have to check all the code. You have to update things, things that used to work

01:03:49 in the software four years ago, now have a slightly different command or whatever, right? And anybody's

01:03:52 been through it will tell you. So we were talking about the layers earlier. When I was going through

01:03:58 that book revision, I wasn't seeking out new clients. I wasn't focused on social media. I wasn't

01:04:04 keeping up with my blog content. That was kind of stealing my extra bandwidth, right? But then that

01:04:10 was over. I came up for air and I recorded a new LinkedIn learning course. But while I was recording

01:04:15 that, I wasn't doing a lot of writing, right? And then more recently, I've been doing this focus on

01:04:21 increasing my followers on LinkedIn, because part of that is learning how to do it, posting every day

01:04:26 and so on. I couldn't have done all three of those things at the same time. If I'd been doing a book

01:04:30 revision, a new course and focusing on social media, it would have killed me. Just the context

01:04:34 switching alone. Yeah, you have to take turns because otherwise you're not going to get through

01:04:38 it. So if somebody has a full time job, it's great to have a half dozen ideas, but you'd better stagger

01:04:43 them or you're not going to make it. You're not going to make it through. Yeah, yeah, absolutely.

01:04:48 All right, Keith, I think we're well over time, but it's been really fun. So let me just ask you the

01:04:52 final two questions and let you get out of here. Sure. All right. So if you're going to write some

01:04:56 code, what editor do you use? Oh, I'm an RStudio fan.

01:05:00 Okay. Yeah, very cool. I've heard good things about RStudio. I've never done anything with it,

01:05:05 obviously doing almost all my stuff in Python. Very cool. And then notable data science package

01:05:10 out there, like a library that people could use that you think is really awesome that you want

01:05:14 to recommend that you enjoy working with. Well, I mentioned ggplot2. Yeah. I'm just a big fan.

01:05:19 Part of it, as you can imagine, is what are things that I incorporate into my courses,

01:05:23 whether they be UC Irvine and so on. And not only do I use ggplot2, I sometimes teach it. So I'm a

01:05:29 fan of that one. Right on. Also, I should mention like some of these ideas, 75% or so that we talked

01:05:35 about came out of your LinkedIn learning course called Side Hustles for Data Scientists. So I'll be

01:05:40 sure to link to that in the show notes as well. So final call to action, people are looking to explore

01:05:46 something different, something more, maybe they don't get to work on exactly what they want to work and

01:05:51 they have this passion and want to do some kind of side hustle. What do you tell them?

01:05:54 Well, you know, I say to, you know, seek out the kinds of things that we've been talking about.

01:05:59 One that didn't come up that I think is a great one is try to find a way to be a technical reviewer

01:06:04 on a book. And then all you have to do is check out your library of technical books, author that you

01:06:10 admire, send them an email and ask them when they're doing their next edition and volunteer to help out.

01:06:15 Yeah. I'm sure they would appreciate it. Actually, it's a lot of work. Awesome.

01:06:19 Yeah. It's hard to find a good technical reviewer and it's not crazy money, but it usually is paid.

01:06:23 But hey, that's a great credibility builder. Not only that, it's a fantastic way of networking with

01:06:28 someone that is probably pretty well known. Yeah. All right. Lots of great advice. Thanks so much for

01:06:33 being on the show. I appreciate it. I enjoyed it. Yeah. Great. Me too. Bye.

01:06:37 This has been another episode of Talk Python to Me. Our guest in this episode was Keith McCormick,

01:06:43 and it's been brought to you by Linode and us over at Talk Python Training. Simplify your infrastructure

01:06:48 and cut your cloud bills in half with Linode's Linux virtual machines. Develop, deploy, and scale your

01:06:53 modern applications faster and easier. Visit talkpython.fm/Linode and click the create free

01:06:59 account button to get started. Want to level up your Python? If you're just getting started, try my Python

01:07:04 jumpstart by building 10 apps course. Or if you're looking for something more advanced, check out our

01:07:10 new async course that digs into all the different types of async programming you can do in Python.

01:07:15 And of course, if you're interested in more than one of these, be sure to check out our everything bundle.

01:07:19 It's like a subscription that never expires. Be sure to subscribe to the show. Open your favorite

01:07:24 podcatcher and search for Python. We should be right at the top. You can also find the iTunes feed

01:07:29 at /itunes, the Google Play feed at /play, and the direct RSS feed at /rss on

01:07:35 talkpython.fm. This is your host, Michael Kennedy. Thanks so much for listening. I really appreciate it.

01:07:41 Now get out there and write some Python code.

01:07:42 Thank you.

01:07:56 Thank you.

01:07:56 Thank you.

01:07:57 Thank you.

01:07:57 Thank you.

01:07:58 Thank you.

01:07:58 Thank you.

01:07:59 you you you Thank you.