0:00

And so I built a computer vision model to detect the school bus going by the house. It sends me a text, and then I know I have about five minutes before. Before the bus is going to pick my daughter up. So we actually are using that. It's running 24/7 I'm getting the texts from AWS, and then after I had that running, it was like, okay, well, what can I do next to make my life better? Hello, and welcome to the Pibytes podcast, where we talk about Python career and mindset. We're your hosts. I'm Julian sequeira. And I am Bob Baldebos. If you're looking to improve your python, your career, and learn the mindset for success, this is the podcast for you. Let's get started. Welcome back, everybody, to the Piewites podcast. This is Bob Baldebols, and I'm here with Kristen Kerrer. Kristen, welcome to the show. Hi. Thank you so much for having me. Yeah, super excited to have you here. Julian would have loved to be here, but as you know, he's in Australia, and it was super hard to align the time zone, especially this time of the year. He's actually traveling, so you'll have to do it with me. But just that, you know, he definitely wanted to be on this one because we're excited to have you here. You're doing so much cool stuff with data in the space, so I'm looking forward to the next 30 minutes to pick your brain, get your background, how you got into it, what you're doing, and share that with our audience. So welcome again. And just to kick off the story, yeah, maybe you can just a broad question, tell us your story, how you got into data science, and what excites you about data science, machine learning, AI. Yeah. So, I mean, starting from the beginning, I got a bachelor's degree in math. And after a couple years, actually, the recession hit, and I had been doing financing stuff for a real estate brokerage, and so I knew I was going to get fired. I decided to hide in academia and got a master's degree in statistics. And so just naturally, that lends to working for data when you got out. So, you know, my future boss came in while I was working in, while I was getting my master's in statistics and was looking for somebody to do econometric time series analysis and forecasting. And so, you know, 2010, right out of grad school, I was, you know, building models with data. And, you know, I've been building models since. Nice. And you currently work at Comed ML as a developer advocate, maybe you can tell us a bit about that transition, how you got into that role, and a bit about the company's mission and what you do day to day. Yeah, sure. Comet's mission is definitely to remove the friction from machine learning, and they do that through an experiment, tracking tool and modeling models in production. And then. So as a developer advocate for that company, I'm constantly evangelizing the product on social media. I'm taking part in conversations where people in MLops are hanging out on the Internet, in places like Reddit. And then I host my own show, the Cool Data project show, where I interview practitioners in MLDL and AI about cool projects that they're working on with a focus on their approach and methodology. Nice. That's a whole lot. Very active in the space. And yeah, I'm definitely going to ask you about some of those projects and also about the content creation. But before getting there, a quick detour, because you've come from R and you transitioned into Python. Right. Another question we had in our preparation is, yeah, how does R compare to Python? What do you really like about Python? And are there any things you're actually missing from R that you're not finding in Python? Well, I think that the answer to that question has, like, evolved rapidly over the last couple years. So my transition itself wasn't pretty. I always say that I wish I had taken more programming courses in university. Just, you know, my math degree was pure math statistics in 2010. They were absolutely teaching everyone R. And there's so, you know, in 2018 is about when I transitioned, the data world is heavily moving towards Python. More of the jobs are offered in Python. Overwhelmingly more of the jobs are offered in Python. And so I actually learned through moocs. And I think that comparing the two, like, in 2018, I had a job that was fully in Python. I was building a project and actually used the RPI two library to do some time series work. I was using the tbats algorithm for a project to determine who had a seasonal usage pattern, but. Right. So I was starting in Python and I was going over to R because the time series capabilities over there were still, you know, just much more robust. And so as time has gone on. Right, like, I was an r shiny person for building my web apps, big fan, and I felt like it was much more intuitive and easy to use compared to dash. But now there's streamlit in Python, which I'd say is almost, you know, even to use than shiny. And so, you know, and ggplot two, which is the big library that everyone uses in R for making their plots. There's now a port of that in Python called plot nine. And so as time is going, both languages are starting to have, like, things that are very comparable in each, which, you know, has made it a lot easier to continue learning Python as well. Yeah. So much dependent kind of, on the tooling and what you needed for the job or for. To get a project done. But is there. Is there anything like, language feature wise that you. You miss from R that you miss? Yeah, yeah. Well, I mean, there comes. So, you know, the grammar of graphics in R is just a really readable way to be writing code that I really like. But at the same time, you know, there's probably libraries in Python that use that now as well. So, yeah, Python, there's so much good coverage in the standard library, but for the more data stuff, the PI PI and the package index is just amazing. Like, there's just a library for everything. Yep, yep. It's like, there's an app for that. Yeah. So talk about projects. You told me the other day about some really cool Python projects you recently have developed. And, yeah, I don't want to put words in your mouth, but one, for example, is the school bus, and I will leave it at that. I will just maybe can mention two or three of those site projects, why you built them and what's cool about them and how they really helped you in learning Python, but also. Yeah. Advancing your career, even. Yeah. So actually, I get my project ideas by walking around my house and deciding what's bothering me and, you know, what I could improve in my life. And so one of those was the school bus. So, you know, it was really difficult because I'd be making an english muffin or something for my kids, and I'm not able to see the bus passing the house. And then it became, did we miss the bus? Is the bus late today? And it actually was like a real point of anxiety and stress in the morning. And we're very lucky that the bus passes our house in the morning, picks somebody else up and comes back and picks my daughter up at the end of the driveway. And so I built a computer vision model to detect the school bus going by the house. It sends me a text, and then I know I have about five minutes before the bus is going to pick my daughter up. So we actually are using that. It's running 24/7 I'm getting the texts from AWS. And then after I had that running, it was like, okay, well, what can I do next to make my life better. And so using similar technology, except now I'm going to be on a raspberry PI pie instead of, you know, using my desktop. But this next one is. And actually I bought a 3d printer and I've 3d printed a case and it holds the little raspberry PI and, like, a $10 camera, and it points at my pill bottles. So every day. This isn't running yet, still in development, but I do have yellow v five running on the Raspberry PI, looking to get that running a little bit better. And I've started annotating data, but I don't have it running end to end yet at all. But so the pictures that I'm annotating is just going to be when I pick up the pill bottle, when it sees my fingers on the pill bottle, it's going to send me a text and it's going to store the data in a database. And so basically that's because there's

many times, but even at 10: 10:18

00 a.m. I'm like, did I take my pills today? And then it's, you know, did I miss my pills? Well, I also don't want to take my pills twice. And so that's been a pain point for me. But also, you know, more for fun. So the text part is very useful to me, but also for funsies, I will build an app that'll have, like, you know, the 30 day adherence rate of, you know, when am I taking my pills? What day am I most likely to forget my pills? I don't want to make it sound like I always forget my pills because I don't, but, you know, that's the next project that I'm working on. Yeah. What's really cool with these projects is that you build up a database of trends, right. And, yeah, at the very start, it's almost no data, but have that running for a year. All of a sudden you see these cool trends and you can make predictions and, yeah, it might be. You might get some mind blowing insights. It's also cool that you combine the hardware with real sensors and stuff and. Yeah, and I think for our audience because we always talk about building apps and you have to build to really get a thorough understanding of the technology. And people have to say what to build. Right. Sometimes they overcomplicate it because you have to build the next Twitter or master don. Not really. Right. Just walk around your house and look at small things, which when you start implementing, can really get out of hand in a good way. Right. So nice. Nice. And then of course, that leads nicely into the next question because you don't just build those things, but then you go actively on social and I think particularly LinkedIn and talk about those projects. Right. So that then leads to naturally to content creation. So I wanted to talk a bit about that content creation process. Yeah. How important you think it is and how do you stay consistent with it and just overall your process. So you build something and do you write a blog post, you produce a video? What is your process? Yeah, so the computer vision project, to detect the bus, specifically that one, has turned into a number of podcast invites. The first article I wrote about it is on my blog. The second one was a guest blog. So I used Roboflow in the project. And so I've made friends with Roboflow during that. And so the second blog was up there. I still have to write the third blog. There's been a couple of things that have gotten in front of that. But for sure, you know, I absolutely look to stretch any project that I create as far as I can, whether. So I also have an email list with 6600 people on it. I'm on TikTok, Instagram. Like, I'm everywhere. I'm on Reddit. I actually do really well on Reddit. And so definitely looking to stretch everything as far as possible. Consistency has been an issue with COVID I actually took like a year off and was really excited to see that when I came back that the algorithm was just as friendly to me as it was before. But, yeah, so, you know, the Reddit one or the LinkedIn one? What? Which algorithm? The Reddit one or the LinkedIn one? All of them. Reddit is beautiful because you don't have to have a following to get mileage on Reddit. So if you do something in computer vision, that's cool and you create a video of it and you share it on Reddit with zero followers, you're going to get traction on that post in a way that you won't on other platforms. Right. But your main channel would be your blog or LinkedIn. LinkedIn. You have quite a big following, right? Yeah. So I have 89,000 on LinkedIn. And yes, that is 100% my main platform. I prioritize, you know, keeping that up above anything else. And is that, I mean, apart from the big following, is that also the most suitable platform, you think, for technical data, Python content? So there's absolutely a huge data community there and the right people that, you know, we'd want to attract to comet is there as well. You know, everyone from all the way from analyst to cutting edge. AI is hanging out on LinkedIn. At least enough of these people. Are you also Mastodon? There's a big flag from Twitter to mastodon these days. Yeah, yeah, yeah. I saw everyone migrating. I just. The way that it's not, you know, you're not going to get organic, reach past the people that are already in the little groups. I don't know much about Mastodon, but it didn't look like something that I wanted to take on in addition to all the platforms I'm currently managing. Yeah, yeah. You got a lot already. Yeah, no, that's. That's inspiring, because we. Apart from big on the technical and the python skills, we also always recommend the people that work with us and our community. Like, you not only have to build, you also have to get it out there. Right? Build up your portfolio so that people can see what skills you have and go a bit more from pushing to pulling in with what we. With which we mean to. You cannot rely recruiters or next opportunity just come to you naturally. Building a day will come that just doesn't happen. Right. As that book, our favorite career book, so good that I cannot ignore you by Cal Newport, says, like, build up career assets. Right. And there's. Definitely. The more you do that, the more opportunities open up. Right? So 100%. Yeah. No, I mean, I can't even express the amount of opportunities that have come to me through LinkedIn that I never would have assumed would have come through. And absolutely, it is paid dividends. I highly suggest that. And I can't tell you about the number of recruiters that are in my mailbox. I hear from them multiple a week. It certainly pays dividends to put some effort into your personal brand. Right? Awesome. But now you're doing that for a long time. Right? How's content creation compare now to when you started? Because I can imagine for people starting out, very scary. There's a lot of imposter syndrome. Right. Um, is. Has it become easier? Are you just pushing through that imposter syndrome these days because of you have done it so many times, or is there still just always that battle? Like, I'm going to put something out there. I'm not sure if it's good enough. Just let's just roll with the punches. Is that still the case, or. Just roll with the punches? Yeah, I. And it's funny as your following gets larger, or I can't speak for anybody else. For me, like, if I do something stupid in front of, like, 500 people, and then I take it down. It's like that just doesn't feel like a lot of people, you know? And I also think for me, so I'm, you know, in 2008, I went to a women in mathematics conference, and they asked, you know, we're sitting in the auditorium, and they say, does anyone ever feel like if your school knew who you really were, what you knew, that, like, they'd still want you there, you know, and everyone was sort. I don't remember exactly how it was worded, but everyone was sort of raising their hand in terms of, I have imposter syndrome, you know, and my hand was up, too. And I feel like the first, you know, five or six years of my career, I really struggled with, oh, I don't know, enough. Oh, I don't know. People probably expect me to know this, but I don't know. I'm turning 40 next birthday, and at this point, like, it just doesn't bother me the way it was. Like, I know what I know. I've been around long enough to see, you know, work with other people and see what they know, too, and understand what their gaps are as well. And, like, we all have things. We know we all have gaps. And I'm just, you know, really comfortable in. In what my gaps are, too. You know, I go into a job interview and, like, you know, I program. But I was saying, I was like, I'm not a software engineer. Like, I know, you know, I know sort of where I feel comfortable, and that translates over to LinkedIn. I don't really worry. You know, I know that what I'm sharing is factually correct because I'm gonna google and make sure that I'm talking about things that I know something about. And, you know, on top of that, like, people may not care about what I share, but that's all right, too, right? Right? Yeah. It's a common sentiment. We hear it a lot, and we still struggle with it as well. Like, before we hit recording, I was telling you about the add fun of Code, Right, which is going pretty okay, but there are some days it completely knocks me out. I'm like, should that happen? And the first thing you said, well, the first thing you replied, or you told me, well, we all have gaps, and that's kind of the way to look at it, right? Like, there's the puzzle skill and there's the app building skill, and they're not necessarily the same, right? So we all will always have gaps, and this is a complex field, and I think the further you get into it, the more comfortable you come with that. And I think you just have to be open about that and I think that's, yeah, the only Remedy. Yeah, well, and getting better at finding the right resources. Right. Like, you know, when you feel comfortable navigating Reddit and Stack OverFlow and, you know, you're seeing what other people are asking and you're also able to get your questions answered as well. The first time somebody tells you that, you should just be, you know, googling your error message. Like it's, you know, there's like some, some learning things, but you eventually get to a place where you're just, you know, we're just all in this together. Exactly, exactly. But we always have that nagging feeling that like I should know more and probably the rest will know a lot more and that's all like self imposed, sorry, go a bit on about the mindset, but that's, this podcast has a lot of mindset in it, so I just want to, definitely wanted to bring that up. So thanks for sharing. Yeah. So final two questions then, any tips for. So our audience is of course heavily Python and as a section of that audience is already working in data or wants to work in data or machine learning AI. So if people are relatively new to that part of Python, of the python ecosystem, what are some tips you have to get up to speed? Are there particular resources or. Yeah, maybe some resources to check out. I mean, we're big on building projects and I'm happy that's what shines true in your approach as well, but yeah. Any tips? Yeah, I don't know. It's hard, right? Because everything's so vast and it's actually easier when somebody is just looking to get into the field and has nothing because then it's like, hey, I'd suggest you go and get, you know, a degree in computer science and, you know, then take some courses as well on MLAI and the, you know, master's degrees that started popping up in like data science and ML around 2016. There was a lot of things lacking with those in terms of, you know, there was a lot of programs that were launched like that didn't have SQL and things that were just like very foundational. And so, you know, I think when you're trying to get up to speed like this is again where being part of a community is very helpful to you. So because the answer is going to be different depending on are you coming from a cs background? Are you coming from a stats background? Have you, are you completely learning mOocs to get into the field? Which any of that's fine, but the advice is going to be different depending on what it is. And there are great programs out there. I learned, my first mooc I took was when I was learning python. I took python for everybody. If you have no python at all, that's a good place to start, but if you have any python at all, that's not for you. And then I also love business science university, has a great python course that it's going to be really python and algorithms as well. And he does a fantastic job with courses. But again, it also depends on, like, exactly where are you, where are you starting? And so I think it goes back to like, hey, like, go start following some of the, you know, Data science, or There's a Reddit, subreddit, learn machine learning, and all of those People are Learning Machine learning. And like, go ask those people, like, hey, if you had X, Y and Z that I already have, what would you do next? You know, just to get more tailored advice. That's right for you? Yeah, that's great. I think the community aspect is very important. We grew our select community to 3000 or more people. And, yeah, the way that people share there and, yeah, the realization that you're not alone and because it can be a pretty lonely journey. Right. So reaching out and do this together with other people makes a big difference. And that's what, that's something I really like about the concert creation being on LinkedIn or wherever. Like, people engage with your stuff and you build up connections. Right. So. And that can be very motivating to, to keep going in this journey. Yeah. Cool. What are you currently reading? Last question, or anything you have read. And that was, that's cool. That you want to share with the audience. Yeah. So the book, I'm holding my page right here. So this is what I'm currently reading. It is the business value of developer relations by Mary Thingvall. So if anyone's thinking of getting into Dev rel, I'd suggest starting with this book. And then the book that I always mention on a podcast is weapons of math destruction by Kathy O'Neill. Because I just think that that is such an important text and that anyone who's looking to get into data should be aware of the different considerations and things that you should be thinking of in terms of ethics and building models that aren't perpetuating bias. Interesting. Okay, cool. They're going to link those below. That last one I definitely want to check out. Yeah, yeah. Cool. Well, thanks for sharing. This has been a lot of insights in just 30 minutes. Any final shout out or things you want to mention? Question for py bytes or me? Yeah, any last things you want to share or ask? No, I just want to thank you so much for having me. It's been so great chatting with you, and, yeah, I'm glad we were able to do this. Yeah, also shout out to Zhu Zheng for connecting us. That should be set and really happy you did. And yeah, it's been a pleasure having you on the show, and thanks again for sharing all these great insights. I'm sure that our audience will appreciate it. All right, thank you. All right, thanks. We hope you enjoyed this episode. To hear more from us, go to Pibyte Friends, that is Pibit es friends, and receive a free gift just for being a friend of the show and to join our thriving slack community of python programmers, go to Pibytes community. That's Pibit es community. We hope to see you there and catch you in the next episode.