Julia Silge - Keynote PyCon Colombia 2025
Julia Silge Engineering manager at Posit PBC, leading development of open-source software for data science in Python, R. Data scientist, expertise in machine learning, text mining. PhD in astrophysics, author of books on data science. Social Networks: Github: https://github.com/juliasilge More About PyCon Colombia at http://www.pycon.co
image: thumbnail.jpg
Transcript#
This transcript was generated automatically and may contain errors.
Hi everyone at PyCon Colombia and Python Colombia, all the communities that are part of the conference of Python in Colombia. Thank you so much for being here for another interview. So I have the pleasure to be here with Julia. So Julia, I have been sharing with you and it's a pleasure and honor to interview. It's a pleasure to be here. Thank you so much for everything you have done and I'm really glad to get to have a conversation.
So can you share with the community a little bit about your background? I know that it has to do with physics, but they will be like, wow. So if you tell us a little bit more about that. Sure, sure. So a very long time ago, I was an astrophysicist. That's my academic background, physics and then astronomy. And then kind of more a medium long time ago, I was a practicing data scientist, a data scientist who worked in organizations, helping to write reports and build apps and train models and do spend really hands-on data science practitioner.
So I had transitioned from academia into data science. And then more recently, I have kind of transitioned again to focus on building tools for data scientists. When I was in kind of that previous time, when I was like, my title would be data scientist, what I actually did was I spent about 80% of my time analyzing data and about 20% of my time building tools. So both open source kinds of packages, and then also tools for internal use, you know, like in my company, can I build a tool that not only I can use, but other people can use. And I really did like the tool building.
So about five years ago, I changed from being someone whose title is data scientist to someone whose title was software engineer. But building the software that I was building was for data scientists to use. And so then I kind of shifted to where now I would say I spend, I don't know, maybe 80% of my time, you know, doing the tool building and then 20% of my time still analyzing data to either, you know, because I need to or for teaching purposes or demo purposes or things like that. So that's a little bit about where I came from and how I kind of got here to where what I do now is I lead a team building free and open source software for Python and R.
Career pivots and dealing with uncertainty
So taking into account that pivot that you have facing in your life, what would you recommend to members of our community that are planning to change careers or have that pivot or they are facing, like, I don't expect the one. Oh, yeah. Well, that happened to me. Actually, if you when I kind of gave that arc, actually, the motivating factor for me to become a data scientist, this was a long time ago. This was maybe this was maybe 10, 11 years ago.
So what motivated me to move from what I was doing into data science is that I was laid off. I was fired from a company. The company that I was working for, I had a role that was like a content development kind of role, like for physics and astronomy and like for higher ed classroom, for like teaching in colleges. So I had a job that was about writing and making content for that kind of thing. The company was acquired by a bigger company and we laid everyone off and then they wanted to hire me back as a contractor. And I thought, oh, this was not fun. I did not like how this went.
So I have experience when like when something is not really your choice, but, you know, you're like, OK, well, I have to figure out now what am I going to do? Like, what am I going to do next? And that was my motivating, the motivating thing that happened to me to say, let me get serious about this thing I've been hearing about called data science. You know, like 10 years ago, this was I was a different situation where I like it was not so much, you know, there were no masters in data science or, you know, like it was kind of before before all that started happening.
So when I think about when I think about that now, I feel like, wow, I'm so happy with how things turned out. But I'm like when you have experienced that, like being fired or laid off, it feels so bad. It feels so bad. And there's so much uncertainty. If I if I were to give advice for someone who is kind of at a transition, maybe not their choice, maybe not their choice, kind of transition, I think I think it's good to get serious about what are your real skills?
What actually do you have that's a bit unique, you know, and what like to have an open mind about what's out there, like what's out there that in terms of opportunities. So I for me that I, you know, I've gone through so much school and had a Ph.D. and I was like, OK, what do I have that's unique? Like I know how to deal with real world messy data because I did that all through all through my time in research and academia. And I like to do very applied work.
And so then I was looking around, OK, what are what can I do out there to try to take those things that are true and unique about me and kind of like find what that what that thing is? I guess another piece of advice is that like these kinds of experiences are so common, like it's very common, like when you go and talk to people, it's very common for someone to have been laid off or, you know, like to have in their career something like being laid off. And it is good to feel to know that it is it is not about you as a professional.
Like probably there was some mismatch or, you know, like I'm not saying I'm not saying it's never about you and what you did, but it's like it's important to be you know what this is not. This is about economic circumstances and the changes in the company and like to not think too much about like this. It says something very bad about me. So I think that would be what I don't think it personally like massive layoffs.
Choosing the right tool for the job
So switching to talks. You went to SciPy 2024. And I saw that the title was like the right tool for the job, right? So like people that are watching, you can watch that is great. So I would like to ask you, like, what is your framework philosophy behind choosing a language or a tool for the job?
Yes. That talk was really interesting to prepare. So a theme of SciPy last year was like scientific computing across multiple programming languages. So like it is a scientific computing with Python conference. But one of the themes was how does Python integrate with other languages like whether that's C or Rust or so I was I think I was invited because I'm I'm mostly known for being an R person. And it's true I love R like I love R. But I actually do use many languages in the work that I do. Like I'm not someone who only ever writes R.
I write a lot of different kind of languages. And I do I think when I think about how do I decide what to pick up, I think I balance a few things. The first thing is what am I best at for a certain task? Like for example, I personally am best at making data visualizations with ggplot2. So I if I need to make a plot, I reach for ggplot2 like I'll write out data from Python and read it. And it's like if I need to make a visualization, I'm just always going to reach for ggplot2. But that's because of like what I am already have a lot of competence in. So so one thing I think about is what am I good at? So I should use the best tool for this job for me.
The other thing I think about is what are the what are the what is the team that I'm working in? What who has what skill? And how do we how can I make the best choice in the context where I'm working? So in that case, I, you know, not now I work on a team that we write, we write mostly TypeScript on the project that we're working now, which is a big data science tool, a new IDE. And you know, we write some Rust and we write some Python. And so in that context, it's like, well, in this case, I'm collaborating with a team. And so I, I don't only think about what I'm good at, I think about what the team needs and is good at.
It's like, okay, if I can come in here and like, okay, we need to change x. And this part of the code is all written in TypeScript. So then I'm like, ah, the best tool obviously is for is here would be TypeScript, right? So it's like, first, what's the best tool for me? In that context, what's the best tool for the team that I'm working in here? And then also, I think what kinds of tools and norms and practices arise in different pockets, you know, like, I think I love to use Python, for example, for like, dealing with API calls, web scraping, because there's like a really nice set of tools that have arisen there. And the I really like if I'm going to interact with like an LM API, I like to use Python for that, because there's some really nice sort of sets of tools that have grown up in that way.
Learning, education, and the era of LLMs
So talking about like, like tools, right, like, you have courses, you have books, right? And I can say and also because of some friends, or some of my friends has read them and taken the courses, that you have like that spark to make like a hard concept really accessible for everybody. So thank you for that. Hey, that's a huge compliment. So so thank you for saying that. Thank you to your friends. Like that's a very big compliment. And I value that someone saying that very much like I feel like that's a big success, you know?
So since you have that spark, I have like, two questions are similar, like where the people that is watching this interview can find your, your resources. And what do you think about like, the education around data, like that, that analytics and yeah, like data science needs to evolve in this time, like in this era we are living in? Well, so folks can find if folks want to find like, like books to maybe read, I do have those links off from first of all, my website is my name, juliasilgi.com. So people can go there to find links to books, interactive courses, video, like screencasts on YouTube, things to watch. So that would be the place to find some of these kinds of materials and resources, blog posts, things like that.
I think it's a great question, like, like, I'm assuming you're kind of talking about in the era of LLMs, in the like time of AI code assistance, like what do people need to learn how to deal, deal with data? I don't, I'll give you my kind of what I think right now. I do, I don't feel I don't have maybe the confidence in what I think now, then maybe I did two years ago, I'm like, I don't know, actually, how are things going to change? Maybe I will be surprised.
I think that there are, I am a huge, huge believer in people learning by doing small things like by doing the small sort of project or a small, like to learn, to learn how to do something. I think it's a mistake for people to think like they have to learn everything about some field and I have to have a total understanding before I can start to do something. I think that the way to learn is by to do a small experiment and then that you are interested in, something that you are actually interested in, like a small and then, and then to grow that and then to say, okay, now I'll do a next thing. So I do, I still think that learning by doing and learning by doing specifically start with something like a small bite, a small chunk, a small piece to, to then that you, you, you are much more, you are much more effective. You are much more able to remember or to gain practical experience by doing that.
I still think that learning by doing and learning by doing specifically start with something like a small bite, a small chunk, a small piece to, to then that you, you, you are much more, you are much more effective. You are much more able to remember or to gain practical experience by doing that.
I, I think that the intersection of how does learning intersect with LLMs is a really big topic and I, um, I think that some things are going to change. For example, one thing that I think is going to change is like how difficult is it if you have competence in one language, how difficult is it actually to transition to another language? Because one thing I think the LLMs are very good at is if you say something in one programming language, it can tell you how to do something similar in another language. Like, so that's an area I think is going to change with, um, the easily accessible LLM code assistance kind of tools.
I think that the baseline computational knowledge, I'm skeptical that that is going to change. Like I think that the still, the way you gain that knowledge is by doing, is by doing. And yeah, I have no problem with, like, I think people using, um, uh, you know, these code assistance kind of tools to generate code and then to run it. Like I think that's, that's, you know, that's no problem, but I think that, um, to be able to evaluate the code that gets generated and to understand what it's doing, I don't think those skills are going away. Like I think the ability to, um, understand what code is doing, the ability to evaluate and review and edit code maybe is more important now than it has ever been, right?
And so I think, uh, my belief is that some things about learning, um, are going to change and some, and I would put like maybe going from one language to another language in that category. But I think some things are not going to change, are not going to change because we still have to gain that, um, experience and skill and, uh, competence in like, how can I know how to run and evaluate and look at and understand what code is doing? It's a good question. And I'll be honest. I'm like, I have uncertainty. Like how is it going to play out?
Building communities
So we're switching a little bit to communities. You have been, you have been part of like R-Ladies and also TidyTest and more things. So can you share a little bit from your ingredients for this recipe of building communities? Ah, great. That's a great question. I, so I, I, I am an organizer of a local meetup in my own city actually, and then have been involved with bigger communities and, um, uh, that are more even global, you know, like even global kinds of communities.
I think that something that I noticed makes a difference is, um, a sense that when I bring my time, my knowledge, my energy, it comes and it, it really, it really multiplies. Like it really, um, like I, I, but I do have to bring myself, I have to bring myself and my, uh, what I know now and, uh, what I'm interested in and like, uh, it takes energy and effort. I mean, I know you're very familiar, you've been spending so much effort and work on this conference, right? It's so much work. It's so, you know, it's so much work and it, so you come and you bring that, but then it, um, you also like, that's, then everyone else does that to some level too. And then what you have there together is, uh, like multiplied.
And I mean, this is kind of, um, I don't know, I don't know about this analogy or what you'll think about this analogy, but it's like, it's like, um, building friendships really because you know, in your, maybe in your non data science life, you're like, oh, I went to work and I'm tired. And yes, I was going to go to dinner with my friend, but I'm very tired. And I'm like, oh, I'm, I can't come tonight. You know, it's like, I'm just going to stay home and scroll on my phone and eat at home and, and it feels easier and it feels more like, um, relaxing maybe. Um, and it's so easy. It's so easy to say like, not tonight, not tonight, I'm not going to come and meet you.
But the, it's like, if you do go and to have the dinner with the friend, um, like, yes, you maybe had to put your shoes back on and put, you know, maybe you decided to put lipstick back on or you're like, oh, okay, I'm going to do it. I'm going to go out. I'm going to see this friend, but you, that, then that, that friend has done it too. And then together you like build in with each other. And I think that the ingredients, it's like, um, it, it really does take, it like takes energy and effort. Um, but then when you're there, it's like the other person also brought their energy and effort. And then it's like, instead of both of you having less, both of you have more, both of you have more.
And I, when I think about, um, when I think about open source communities, I like, I think that's an analogy that works because it's, it, it, it took something for you to show up. I mean, it took a lot for you to show up right here. It took something for me to get here. Right. But, um, when we, it's not like we both now have less, it's like now this whole, this whole, everyone who's here at PyCon Colombia, like we all have more.
When I think about open source communities, I like, I think that's an analogy that works because it's, it, it, it took something for you to show up. But, um, when we, it's not like we both now have less, it's like now this whole, this whole, everyone who's here at PyCon Colombia, like we all have more.
Advice for women in technical careers
So you, I can see you as a role model, right? I would like you to share with more girls and women that may see this interview because you have been loud, vocal advocate for women in data science. Like what would you recommend to them? Like what would be a message for them to keep pursuing that leadership and to be leaders and to be part of that science area and keep doing work with this technology?
Yeah. So I'm going to say something kind of specific, actually kind of specific. And I don't know that everyone would agree with what I'm about to say, but this is my, this is something I think I've observed. So I think, you know, girls are socialized to really like to take care of other people, to be nurturing, to be, to value connection. And so we grow up that way often. And so like often, often women are actually really good at building connection, are really good at that connect, like understanding what someone needs and kind of helping them to have it.
And how that can play out in technical careers is that like, if you're going to get really concrete, like often women are identified as like, Ooh, that person would be so good as a manager. You know, like that person would be so good. She's so good with people. We should have her start to do maybe more like, Oh, we can have her do like the less technical work because she's very good at it. And then it's like, Oh, well, she's, she's less technical, you know, like, is she so technical?
So I think I, an advice I have is to, um, to not step into roles that are explicitly about, um, the quote, less technical work before you really want to, or before you're really ready. I like, I think I do. I have advice. Like, I think it's a little, a little more dangerous for a woman to, um, maybe stop being hands on with coding work too early. Like maybe it's a little, it's a bit, a little bit risky because then you end up, um, uh, spending more of your time doing maybe, uh, organizational work and yes, yes, yes. You would send up spending more time on that. And then in the longterm, it can be a little limiting to your career options.
Now I think that work is super valuable. I'm not saying it's not important work, but I think, I think if I had advice for especially early career women, it would be, um, it would be, if you, if you like technical work, if you like data science, coding, um, be, stay on a track that is a more, a more traditional technical track for like a little longer, maybe like be, don't too early kind of go to these roles that are like, um, I don't know, I don't know, like, like manager or, um, maybe product manager or DevRel, you know, like, like just be, just be thoughtful, be aware that, um, depending on how people perceive you, like, like you might want to wait a little bit to make some kind of those choices.
And that I want to emphasize is not because I think that work is, is not important. It 100% is important. We have to have people doing DevRel. We have to have people who are the managers of our teams. You know, I am now a manager, right? But I actually was nervous about stepping into this, like, oh, she's less technical kind of roles because of what it means for the long-term. Like, I think it can provide, it can be, it can limit your long-term options by going out of the technical role too early.
So I know that's very concrete kind of like specific kind of career advice, but I, um, I base this on, I feel like what I observed happened. And it is true that you end up people who are like raised as girls and adult and then now young women or adult women, like we're, we, we are, we tend to be, um, very good at some of these skills. And so people are like, ah, come, come do this thing, come do this thing. And it's like, well, maybe, maybe wait, maybe wait a few years, maybe wait a few years.
So, because I do think that there's, um, it's really exciting to see women, uh, step into roles that are really aligned with their interests. And I would just want people like, I want, I, you know, I want men and women, but I, including women, like I want women to be able to, um, uh, see roles that they have as like the fit for them, you know, as like, ah, yes, this is what I want to be doing. And so keeping your options open, I think can be, um, a really good idea.
Thank you, Julia. Thank you so much. Thank you so much. I really appreciate your time. Thank you for being here. Thank you for sharing the, all of these advices. And yeah, it was an honor to interview. Uh, it's a pleasure to speak with you as well. Thank you so much. Thank you. See you in the, in the next one.