PyLadies Panel: AI Skills & Careers

Speaker 1 [00:08]

Hi everybody, I guess everybody is still full of energy, it's just the beginning of the day. I have my questions on a sign which you can all do in the Feminist AI LAN party place, so you can do signs. I didn't want to have my laptop here, so yeah. So today it's my pleasure to host this exchange on careers and skills that we need in this age of everything AI. I have three guests today, and I'm giving them the microphone to introduce themselves. Of course, everybody knows them already. They are all either giving keynotes, talks at the conference. I hope you've seen them or you are going to see them. And so maybe you introduce yourself with something that the audience might not easily know about you. Sorry. Plot twist.

Speaker 2 [01:06]

Okay, I will start. Hi, everyone. My name is Anastasia Kravdina. As you already said, I had my talk yesterday, so I'm completely relaxed today. I'm a technical lead in a big enterprise company, and in my free time, I'm also a mentor in ML and AI, when I'm helping people to, you know, to have strategic investment or best investment on their time and have a fulfilling career. So I'm happy to talk about this more tonight.

Speaker 3 [01:37]

Well, I'm Guadalupe Cañas Herrera, everybody calls me Guada, easier, with the old-fashioned Spanish name. I was the keynote speaker this morning hyping you up with the space science and with cosmology with the Euclid mission. Something that many people do not know, but every single Euclidean actually knows, because I'm really insisting, is that I'm a sport instructor, I'm a Zumba instructor. So I teach Zumba three times per week just to combat all the stress around. And that is a beautiful connection as well with the fact that I have a bachelor's in music and piano. So yeah, plot twist.

Speaker 1 [02:13]

plot twist very very nice i'm also doing teaching yoga three times a week so also later today before the social event if anyone wants to practice yoga come to the lan party

Speaker 4 [02:26]

Hi, I'm Jesper. My pronouns are they, them. I just gave my talk where I had a who am I slide and I'm like, I literally said everything. I'm really neurodivergent and my ADHD is really bad today. So I hope I can listen long enough to actually answer. So neurodivergence representation here.

Speaker 1 [02:50]

Who else is neurodivergent in the room? Like, you don't need to, actually, nobody's filming it, so it's fine. Sorry. Okay, so, first question that we have for our panelists today is, in the past couple of years, the tech industry has been changing with the rapid evolution of AI technology. How has this affected how you work?

Speaker 4 [03:15]

I'm a scientist for machine learning, so it has completely changed what I do. I originally studied geophysics and now I'm completely in a different field applying big models, big computers to weather.

Speaker 3 [03:34]

It has changed almost everything. Well, you already have seen from this morning the large amount of data that we have to process. We could only do this now with the new machine learning algorithms and artificial intelligence. Otherwise, there is no sufficient paid scientist or even scientist in astronomy to do everything that we actually need to do. So it has been a game changer and definitely like a window opener to be able to draw way more conclusions than we were able to do before.

Speaker 2 [04:08]

Yes, I think it's the same for me I can only relate to this I guess if someone would tell me three years ago that I will work in NLP. I would laugh at them. But now right My team is building solution with email and it's many of the solution are LLM based, right? So we're kind of expert in this by now and also not only what we work on but also how we work on right So nowadays you have all these nice plugins which help you to code And to get answers for your questions faster and all other tools, right, which in the past you could barely imagine yourself using. Nowadays it's kind of your normal life.

Speaker 1 [04:49]

Does anybody feel stressed out about this development, the quickness of how fast things are changing, or you're more like, oh, yeah, that's cool, I can do all the new cool things all of a sudden?

Speaker 3 [05:05]

Well, I think that in academia, of course, there are many subfields that investigate actively artificial intelligence and machine learning algorithms. So for them, I think that it has been a joy. For those who have to adopt, because as I said, we pretty much didn't have any other choice if we wanted to face the challenges of analyzing all the amount of data that we have to analyze, it came with some concerns. about how the machine learning algorithms are actually being trained, how they are learning, and how much we could rely on them, on the precisions of the analysis. So yeah, we're getting there, but definitely it's not one thing from one day to another.

Speaker 1 [05:56]

And now I have a very fun question. My favorite one. No, it's the second favorite one. So it feels like many organizations are trying to become AI-driven before they have truly become data-driven. So first, do you think that's good? That's a rhetorical question that I copied from Ines from the previous panel. But that's not the question for today. How do you feel like this is affecting hiring needs? And follow-up question, maybe whichever you want to answer. What skills do people need now to adapt to this? Who wants to go?

Speaker 2 [06:41]

Maybe I can start. I believe nowadays language becomes an interface to many things, right? So essentially, instead of Googling stuff, you chat GPT stuff. Therefore, I personally believe that writing is actually an extremely valuable skill, right? So on one hand, it sounds silly because why do you need to write anything? You can just share your thought with chat GPT and it will create a nice piece of text. But this is exactly the point. So the better you can describe yourself in words, the better the prompt you can write and get what you want with LLM help for example and this is maybe not so it does not sound technical at all but I do believe that everyone who is serious about tech should become a better writer

Speaker 4 [07:26]

I think it also, like, on the hiring situation more, it affects churn. You really have to be careful when you're in an interview to also interview the company. Like, if it's a company that now wants to do AI because they talk to McKinsey and they have only PDFs or whatever, you're probably going to get burnt out. And if they're hiring the second person, you probably want to talk to the first person why they quit. It's like a lot of companies are trying to like jump the whole data situation. So that also leads to you need to have foundations like writing is a great foundation, but like data skills, especially if you're going into a company, you're not going to use LLMs or fancy stuff. Probably you'll do a decision tree or a random forest after you've spent months of cleaning data Getting data into any type of sequel light or any type of database if we're being realistic

Speaker 3 [08:37]

I mean, I would like to advocate about the fairness and ethics that we still have in the pure academic world of cosmology. I think that we are still pretty much data-driven. I think that we still are in the phase that we open up the data and start playing with it to investigate what it has to tell us, but let's be honest, machine learning gets you funding grants and research grants. And I think it's worth mentioning as well that sometimes I wouldn't say that it is not overselling. Just make it a slightly more hyped so that, yeah, okay, maybe in reality you just need a linear regression, but okay, let's use Gaussian processes instead just to see if we could potentially investigate and stretch it further the fact that, okay, so machine learning is actually all hyped these days. And I think that this is really, really dangerous in my environment. I'm not seeing that that much, but it is a fact, just data statistics are there, like the latest funding, big funding projects in astronomy, somehow they have been related as well with some kind of machine learning or using new data science techniques. So yeah, I think that it is pretty much colonizing different aspects also in the academic world.

Speaker 2 [10:01]

Can I add something? I think what you shared is an interesting trend. I have been in the research, and I can absolutely confirm if you have ML in your grant, the chances that you will get the funding are much higher. And I think there are good things about this. It's great that we can train more people in ML domain, right? Because if you promise to use ML, you're kind of obliged that people in your team who do research will actually use it. But I also see a similar trend in enterprise where I'm working right now, that every project you start must have something to do with artificial intelligence, right? So you kind of, there is no way out of this. It has to be this way, which might be also good or bad, right? So everyone is very excited. We want to use the latest technology. But sometimes you ask yourself, okay, maybe boosted decision tree will be enough, right? So you don't need to plug LLM in absolutely every single thing, especially if your data is tabular data, right? And I think here we as an expert in this field have responsibility to actually tell people, look, I know you're absolutely crazy about AI and chat GPT, but let's just sit down and think if it's actually worth it to use it here.

Speaker 4 [11:11]

Maybe just to add a joke answer to that, a lot of people publish that they have a fantastic transformer deep neural network architecture and then deploy the linear regression or the random forest because you actually have to run it and GPUs are expensive.

Speaker 1 [11:33]

So, coming back to, well, continuing with the AI hype, I have a question about learning. So, we have a lot of excitement about AI, also not just at the C-level, but also anybody who's currently learning anything. And it's very appealing to just say, why should I code it when Claude can code it better? Anyway, some people might say that. So, and you're doing a lot of educational material, and how can you, what do you think people can do to balance how they use AI tools with still doing independent learning, like, you know, using their brains?

Speaker 4 [12:14]

I think so it's on multiple tiers because like if you're learning something like a skill like programming Python, we actually got training from Florian Bruhen, the maintainer of PyTest. And at some point I realized I have to switch off Copilot during the training because I'm not learning right now. So in that case, just literally switch it off because you're supposed to learn. you need to learn like the syntax and what it's doing and you need to form the connections in your brain yourself you have to do the hard work to learn it and that is a safe environment like if you have to do it on the job and then something doesn't work because the ai agent wrote something and you're just sitting there that's super stressful so use the time now to like save yourself pain in the future and it's an it's an investment in the future but if you're like learning about AI itself that can also be really overwhelming like all the talks seem to be about AI agents and before that they were about RAG and honestly I'm a scientist for machine learning I'm working in this field right now I know that what RAG is, I've never deployed RAG because why would I and I've literally heard about AI agents for the first time in the preparation for this panel where I'm like do I have to look up what they actually are Because, like, it's just an LLM in a for loop with some engineering around it, if we're being honest. And it's, like, the hype can be really big. And social media really, every influencer is telling you, oh, this solution, that solution, you have to do this to get a job. And it's, like, if you get the fundamentals right, if you know your machine learning, if you know, like, you're prompting maybe, then you're probably fine. If you then read up on the agents, on RAG, on all those, you're like, oh, so we just inject the response from a prompt into a larger prompt. Cool. So then it's really important to see the hype, see what's happening, but not get swept away with it. Because if you have strong fundamentals, you will probably be fine.

Speaker 3 [14:36]

Yeah, this is an interesting point, actually. It just makes me self-reflect right in the spot. Okay, sweaty. No, but I think that this is a question that not necessarily is being applied yet, at least in my field, yet. Maybe we are there, like, in a few years, and seeing how everything is going so fast, maybe it will be next year. But so far, we are rooting in the scientific problem that we have. And then because we cannot solve it with what we call in our colloquial language classical tools, we have to investigate how to incorporate something new. But it's more based on the need and then investigating what might be the best artificial intelligence algorithm or what machine learning packages we have out there that we could potentially use, the way that it comes and it is incorporated into solving the scientific problem. So, for instance, what I was telling about this morning, so we have a package that does statistics to compare different theoretical models against astronomical data, because doing the analysis itself is very time-consuming. Therefore, the problem, we want to have results relatively soon, not in the next century. We need to accelerate it. And then because we need to accelerate, we investigate what would be the best ways of accelerating the pipeline. But it wasn't just, okay, because this is something that is receiving some attention or is hyped, let's incorporate it and let's see what happens. But maybe this is the way that we will go. Maybe next year, who knows. But I don't know, I feel relieved at least. I'm not there yet.

Speaker 1 [16:30]

So, I have a question for you, Anastasia. From your experience in mentoring people, how do you feel that AI is reshaping career opportunities? And maybe if you have some practical suggestions, strategies that you would recommend for people thriving in this landscape at the moment.

Speaker 2 [16:51]

Right. It's a great question, by the way. Yeah, I think AI is changing what people do, not just in tech, but also outside the tech. And everyone who is approaching me, regardless how familiar we are with this topic, I always suggest to look at this and just understand how it works and what it can do. And also accept that most probably what you learn today will be outdated very quickly because even LLM you work with today is the worst LLM you're working with in your lifetime. right if it's keep improving like this and this is a nice thing to think about right I think as soon as you're working in tech you need to accept that what companies are looking for and this wish list or shopping list getting longer and longer and yes it's still a shopping list it's kind of nice to have and many of the things they can neglect if you're a really great candidate But it's true that compared to working in data field five years ago, right now the things you need to bring to the table are much, much larger and much broader, right? So it's all about not just focusing on your little part and knowing how to train neural network, but also knowing, okay, how to formulate the problem, how to decide whenever or not you need neural network, right? So, or any other fancy tool, or if you can solve it in much more classical way, right? So it still exists. and also how you bring it from zero to market, right? So from something which runs locally, which can be running for your users, right? So it's really like this end-to-end thinking, right? And of course, it's a big change. It's a huge change. And I think for people who are entering this field right now, of course, it's a huge threshold which keeps growing.

Speaker 1 [18:43]

I heard someone saying that maybe people shouldn't study computer science anymore.

Speaker 2 [18:50]

No, I disagree with this, right. I think you do need, and here I'm just beside, but you do need to know basics, right, because yes, LLM can write a nice piece of code for you, which looks nice, right, but if you don't have background on this, you won't be able to judge. If it's running, it does not mean it's a good code, right. And this all kind of system thinking, this is also something I believe for many years still be on us, right. how to think what exactly do we want, not just what the results should look like, but also how to build this in the most efficient way. And I think computer science is kind of a very fundamental thing everyone working in tech should be aware at least about.

Speaker 1 [19:36]

Does anybody want to add anything?

Speaker 4 [19:40]

In my talk earlier today, I had an example where I asked ChatGPT or Claude, Claude is the better model, to write a bash script for me to basically make multiple repos into a monorepo. And it was a massive bash script. And in the end, you should just use Git subtrees. And it's like a five-line script. And it's like the LLM is only as good as the LLM is. And you, in the end, like, you have to have this knowledge. And, yeah, I often have it that I run into the context limits of LLMs. Like, the code just gets too complicated, too long, especially because LLMs also write overly complicated code. And you usually have to be like, no, simplify this. And, like, also, like, get it down. and yeah like I don't know I think I'm obviously repeating myself that we need the fundamentals to even understand this but also to take that opportunity to learn where to put things you cannot become a senior by vibe coding yourself into a corner I think like to what you were kind of scratching on there like software architecture is really where you become like a senior computer scientist where you understand, oh, I'm building the core of this right now, but I'm leaving space for those pieces that still have to go. And I personally, I learned that through projects like in science, working on smaller things and then like overgeneralizing them and realizing this was really stupid. Those are use cases that will never happen. and that way I could kind of gauge where in future projects I actually found like the sweet spot where I left space for other things but where I made it specific enough to actually run.

Speaker 1 [21:44]

God, I have a space question. Not really. Well, the word space is inside. In your work, you hinted this earlier in one of your answers, but I wanted to ask you, how do you juggle pushing for innovation with the need to work with tools you rely on? And here comes the space part. Is there any space for code written by agents?

Speaker 3 [22:15]

You know that whatever I say, then it will be used against me once I'm back to my community. No, of course not. Well, it's definitely a challenge, especially because I think that we could address it, like, on two different sides. So one thing is the full astro space sector. If you're building a project, like, with few collaborators that, of course, you know that They are definitely trustworthy because they are, well, fantastic professionals and you know that they know how to code, how to use AI algorithms and different packages. You have already built trust. I think that we are also in the process that many things are already accepted within the community, so you don't question them anymore. Now, that is different when you work for a big collaboration, that this is my case. At the end, you are meant to deliver results that are going to be paving or are going to set a new standard in the way that we are building knowledge, for instance, these days in cosmology. So you really need to build up trust that the machinery that you're using, in some cases, to speed up what you have of classical pipelines is the correct one. So we definitely spend tons of time to make validations. And with validations, it's like, okay, if I'm training a neural network because I need to predict, yeah, the solutions of high order differential equations that takes forever to be solved numerically, I need to be sure that the result that they are meant to be are within, I don't know, whatever decimal agreement with those provided by the neural network. Because otherwise I might be able to reclaim a new physics by the fact that I misinterpreted or I over-trained a neural network and definitely this is something that you do not want and you don't want to have the responsibility of a big collaboration. Now agents, it's getting like on fashion these days and mostly for being able to train new students. You take a code. Yeah, so you take one of the classical pipelines that were written in Fortran, maybe with some Python wrapper on top, and you take on a student that actually, that was my experience, like six years ago when I started my PhD, that I inherited some of these codes that had really poor documentation, people, like it was tough, and you don't know how you're going to start. So for instance, yeah, natural language processing and agents are being a game changer for now that some of these codes have like really solid documentation, get them trained. So you could learn by asking, hey, how do you actually produce and how do you output this? And then you get the answer. So it is a way of making like the user to go a little bit more, I don't know, like if If you will have like a mentor or a tutor that it will guide you through the process. So yeah, definitely this is something that I think is going to be high, yeah, skyrocketing like in the next year, next couple of years, which most of the coach.

Speaker 1 [25:35]

I suppose that's also relevant for industry, where people in the past didn't like to document their own code. Currently.

Speaker 2 [25:44]

You never have time for this, right? Of course. So, like, you must deliver something and move to the next thing, right? Yeah, it's a bad habit, I must say. So I always encourage people to document the things they did, even for themselves in the future, right? Because if you come to something half a year later, you will be very thankful that you wrote this nice piece of documentation as a reminder.

Speaker 1 [26:09]

I have a question, the next one is, so when I told someone about running this panel, their reaction was, there is no good career advice that people can give anymore. So would someone here want to share any timeless career advice, something that worked in the past, maybe still working now for you, something that helped you, or new career advice that you feel like is especially relevant these days? Unionize. Hmm? Unionize?

Speaker 4 [26:46]

In a way, I'm sorry to go that way, but AI is a way to replace workers, right? So it's a way to do things much faster. And back in the day, it was the Luddites that were replaced by dangerous machines that would suck children in and mangle them. And today we have AI that steals intellectual property in a lot of places, especially from proficient writers and the people that have built these skills. So seeing that you do the work, you get the skills, and you get to working with this is important. But also collective action is more important than ever because we're seeing the rise of fascism through VC funding at the moment and we have to be like super careful where this world is going and like being a collective, working together and sharing this knowledge as well for free and being like, hey, we have this open source thing is a way to counteract this entire movement. I'm sorry to go full on leftist on this.

Speaker 2 [28:01]

Right, I absolutely agree with this. I believe communication skills are crucial skills and it was like relevant ten years ago And it's relevant today And I believe in the future will be relevant more than ever because in the end of the day you work with people, right? It might be in research. It might be in the company, but you always work with people and if you are good communicator you can solve like half of the problems which simply won't erase if because you know how it is normally Normally, most of the problems come up because of miscommunication, right? Or because of bad communication and so on. Also, I think being good in your job and what you do is extremely relevant. It might be that in five years from now, you will have absolutely different job and different responsibilities. But knowing how to do good what you do right now will help you for sure in the future. And that's also absolutely relevant for any role.

Speaker 3 [28:57]

This is really interesting and striking to get to hear what you will go first. Because as I was hearing back the question, for me it would be, I think that there is an advice that I received that I will pass to my students, which is make yourself accounted or attached to a good mentor. And a good mentor doesn't need to be your PhD advisor, a professor. It could be someone that you get to talk to, that you feel that communication is flowing, that it could give you like some piece of advice when you need it, but also that could give you like a piece of advice when you think that you are not needing it. And I think that it is really relevant, at least these days, in academia. Like I wouldn't have managed to be where I am now if it were not, of course, by the support of many people, by exploiting opportunities, but at the end of the day, having the support of some colleagues that I consider my mentors. And I never had any problem to just type a question and say, hey, I don't understand this very basic Python thing. Explain it to me. And I think that that is extraordinarily essential. And make sure if there is still, like, some people around here, like in academia or not necessarily only in astronomy, but yeah, I don't know, thinking about some, a career still that it is connected with some research somehow to make sure that you still have like a figure like this one because it could get tough, especially if you're thinking about like a career change or something like that. And as I said, not necessarily have to be someone from your same field. It could be someone that you find through a mentor network. and keep it in time, which is the challenge. But it has to be there. So that would be my advice, I would say.

Speaker 1 [30:56]

Thank you. So before we go to the last question, I want to announce two things. So first of all, we are going to have quite a lot of time for audience questions. So if you want the Slido link, if you want to have questions to ask to the audience, there's somewhere on Discord or there's an everybody should know how to get to the Slido link for this panel, I'm sure. Otherwise there's already some questions. There's a QR code on the wall over there. So ask your questions. And the second part is on what Guada said, is who here feels like they need a mentor at the moment? Okay. And who here is not mentoring anyone at the moment? Nice. So please look, the people who raised their hand that they need a mentor, look at the the ones that raise their hand that they can mentor and they have capacity. Attack them after this panel with your requests. In a nice way, anyway. So my last question in this panel is about the future, trying to end on a positive note. So yeah, looking at the future, anybody who's been on LinkedIn maybe has seen this post about how maybe next year half of the code is going to be written by Claude, the year after 75% of the code out there is going to be written by Claude, and the year after the next job is going to be the person who cleans the code written by Claude. So that's going to be the hype job of the future, code cleaner. And I'm not sure if anybody can actually challenge this opinion, but I'm asking our guests if there are any other exciting things you see ahead in tech.

Speaker 2 [33:03]

I can go first, of course. First of all, I absolutely hate all the zoo of titles we created in the data field, right? And sometimes some companies have their own definition. Let's say data scientist is a great example. What you do is a data scientist highly correlated with what your company does, right? So in one company, you can do a simple data analysis and Bayesian statistic. In another company, you work with LLMs, right? And you're both data scientists somehow. As much as I hate this, I also see some trend going into convergence of things, right? So we have a lot of engineers, data engineers, machine learning engineers, AI engineers. And I hope one day we won't have these flavors because these people also do many things. You essentially do what is needed, right? And not what your title suggests you're supposed to be doing. So I believe at some point we will converge in just software engineer, right? So we will be all software engineers with various flavors. Why not? Regarding new activities we will do with, unfortunately I can't predict, I think we are too early in this journey to say for sure. Whatever activities we will need to do, maybe code cleaner will be the thing, who knows, right? But I do believe that you should be open, right? So what I said earlier, you should do your current job good, to be appreciated, but also be open and try new things. Just try it for yourself for fun to see how far you can go. And another opportunity which AI brings to all of us is all these wonderful tools we have to do the things we could not imagine doing ourselves, right? So everyone can create a nice picture or write a song, right? And it's something which was not possible. And I do believe in these tools and always encourage people to explore them and also look at their users, how they use it, because quite often we are a bit biased because, ah, I know everything about transformers. But you can go into the room with graphical designers who teach how to use mid-journey, and you're absolutely amazed which tricks they found in this.

Speaker 3 [35:14]

Yeah, so I think that, yeah, so up to my mind in my field where I'm seeing is that these days it's already very complicated, and I guess that it's going to be extremely complicated, nearly impossible in the near future, just to say that you're a plain theoretician. Okay, so we are in the era of data, and we need to build some data analysis skills. Maybe last decade you could still survive in this world and build up a career fully based on writing equations on a blackboard. Not anymore. And I've already seen that. Like in my case, I always loved coding. So for me, the transition towards more and more and more coding and being exposed to new algorithms and machineries and keeping up with the hype was a natural step. But many of my colleagues that really generally started as theoretical physicists, like really pure mathematicians, this is gone. So I think that effectively in the incoming years, it will be really hard to find someone within my particular field in cosmology that really doesn't know how to make even a plot or matlod leap. But we still have these cases, and of course you can build a career out of that, which is perfectly fine and honestly really admirable. Now with respect to where we are hitting, I think that definitely, well, for our case, going into autodifferentiation, Jack's framework is really something that it is hitting extremely hard because of the flexibility that it imposes, well not imposes, actually offers to our activities. So I think that definitely we are going to see like not a revolution but clearly a change in the way that we are writing code these days, at least within my field. Now what it will be looking like for jobs application where it will be some of the requirements in the next psd or postdoc positions it is really hard to say because now if i think what they were requesting two years ago it's completely obsolete so i don't want to make any predictions on here

Speaker 4 [37:36]

Yeah, I think I have a little bit of a shift, actually, because I think there's basically two levels of jobs that are coming up. Because one ties into what we talked about before. Everyone's like, oh, all the new things are happening and we have to run, run, run, run. And it can be very, like, it's a siren's call, like, running behind AI agents now and whatever comes in five months, whatever. You need someone that is good with strategy, so someone that can have long-term vision, and you can be a person that just has a very good intake of information, distilling that into some kind of like long-term vision what happens within the next couple of months maybe years because like we've seen transformers in 2017 everyone agreed that this is a paradigm shift as a as a technology and you could see that probably llms are coming down the line and with llms and text understanding there was going to be ai hype coming so if you are good at that then building strategy is good but not everyone is right there's few people that are that visionary in a sense but like this is extremely valuable but also and I'm patting my own shoulder there there's always the need for helpers like there's the Mr. Rogers quote like always look for the helpers like being the support character that enables others that enable science that enable engineers, like someone that fixes your CI, someone that fixes the things that no one else wants to touch, so basically becoming the not the Tatortreiniger, but the plumber, in a sense. The janitor. Maybe more plumbing, because you have to put the pipes together and, but like I think every metaphor of a physical labor job that we have at the moment is also going to become more valid in tech or in techie jobs. I would also encourage you to look outside of classical tech because we both work at coordinated organizations and those jobs are actually really cool but no one knows about them and it's like I predict the weather, you do space and you're more classical work but look outside of the classic Google meta and all that, there's some really, really nice jobs out there where you can do applied work. And so I feel like what you were saying with application is really where a lot of gold lies for careers and building something that you can actually like.

Speaker 1 [40:37]

So I hear a little bit of less specialization and more generalization and building the skill for that Which takes you through special special specializing at some point, but I?

Speaker 4 [40:48]

I always talk about T shapes. So you have to have a deep specialization in something that can be AI, but maybe it should be something more fundamental like software engineering, physics, something like that. And this is my bias. I'm a geophysicist that is good at coding. But, like, having that as a background that you can draw off with the T where you have some knowledge across, like, you're a fairly good writer, you're a fairly good, like, theoreticist, you can solve a differential equation at least, then you can draw from this deep well of knowledge and basically have this inform across a generality of knowledge. So I would more go towards that and then people make all kinds of stuff out of it. But like start with the tea.

Speaker 2 [41:42]

Maybe just to add to this, this long line in tea, it can be like a domain knowledge, right? And the domain might be what you said, it might be physics, it might be user understanding, it might be the domain your company is working in, right? So it might be a project management, right? So it does not have to be a technical knowledge anymore, but I believe like everyone involved in this somehow must have some understanding how it works, right? and this is where this short part of tea comes into the game.

Speaker 1 [42:17]

Thank you. So now we have time for questions from the audience. It's going to be exciting.

Speaker 4 [42:25]

Chihuahua cards.

Speaker 1 [42:27]

About cats?

Speaker 4 [42:28]

No, the wild cards the wild

Speaker 1 [42:28]

No. The wild cards. Wild cards, yeah.

Speaker 5 [42:31]

So the first question is about mentoring because it was one of your suggestions and the question is How to find a good mentor especially when you are not surrounded by data scientists, maybe you can open this up a bit

Speaker 2 [42:50]

I can give a couple of suggestions. I'm a mentor myself, right? I just had a talk yesterday about mentoring where I mentioned that we have way too little mentors. Therefore, I like what you suggested. Everyone who is not mentoring anyone yet, please volunteer for this. I think it's great. There are also many platforms out there, right? So if you just Google mentoring in email, you will get like thousands of various paid and unpaid places where you can look for the mentor. but what is important is match between you and mentor, right, so you need to relate to this person, there must be something in their experience where you say, okay I believe this person can share something valuable, like people who are coming from research and now working in industry, a fantastic mentor for people who are coming from research as well because they know very well both universes and can help you to connect yourself to the next universe you're heading into, right So I would say this connection, so when you read the mental profile, you need to see why you might be connected to this person.

Speaker 5 [43:54]

There is no limit. You can have more than one mantra.

Speaker 2 [43:56]

Absolutely, this is also, you can also have mentors for various things, right, so more technical, maybe more product oriented and so on.

Speaker 5 [43:57]

Absolutely.

Speaker 1 [44:04]

and so on. I would like to add something to this. I think a lot of people who are getting into mentoring other people are very scared that they have to be an expert and you don't. Honestly, anybody can mentor someone. Sometimes it's really valuable just to mentor someone and you can hear yourself talking out loud those things that somehow you wish someone else had told you just a month ago. So you don't have to be 10 years older or ten jobs further than the person you're mentoring you just need to be even maybe just learned one extra thing than that person and so anybody can mentor.

Speaker 2 [44:44]

And it actually helps if you are not too far, because as a mentor you relate better to this person. Because if you have been in his shoes, let's say, a few years ago, it's much easier for you to understand the situation.

Speaker 5 [45:01]

Following up on the mentoring topic actually, do you have a role model and what are your personal experiences maybe from good examples of mentors?

Speaker 2 [45:17]

I was fortunate enough to have people in an organization I was working in, great people. I would not mention their name because you don't name these people. They are not famous outside of the field of their company. But this was the group. I mean, you usually notice these people, right? So it's not just they are nice, but they are also very professional. And there is something you like about them, right? And you see how they work, right? So you don't need to follow how they work exactly, but it's definitely something you can relate to. And from our field, I think there are also a couple of voices. For example, head of AI from Meta. I mean, Meta is evil. No one wants to think about this. But he's a very reasonable guy. Everything that he's saying about AI, I totally agree with, right? So, well, maybe not with everything. It's not like I follow him and read every day what he writes. But I do believe that even in this evil company, you can find the people who are quite good.

Speaker 3 [46:17]

Yeah, so if I might comment on this, actually, I think that it is important to understand as well what a role model for ourselves means, because at the end, we tend to not idealize, but to admire the people that intrinsically, even though you are not realizing, represents the principles that you aligned most with. And I think that something that it works in this regard is to understand what are your principles and what is basically maybe your motors or what it is driving you forward in your career path. So you could identify people that maybe, as you were saying, like not necessarily are the most famous or like admirable, actually the people that I admire the most and I would definitely consider them role models to follow would be maybe some colleagues of the Euclid Consortium in my case that absolutely no one knows anything about, but they are representing like really intrinsic values of altruism, like doing good science, hard working, discipline, and being there always willing to help and for the good of science. And it might be like a view a little bit, I don't know, too optimistic from what the science represents, but in some sense it motivates me as well. So yeah.

Speaker 5 [47:40]

Very nice suggestions.

Speaker 1 [47:42]

Wait, maybe, yes, do you have a promo?

Speaker 4 [47:48]

Mine are a little bit different because they're much closer to me. There's people in Python community, for example, actually from PyLadies here who did the panel last year. And I'm mostly inspired by actions and by principles as well. If you chase fame, you will fall flat because I think most people in this room at one point thought Elon Musk was doing really good stuff in the world. And never chase fame. Thank you.

Speaker 5 [48:21]

Then, continuing with something slightly different, it was several times mentioned, learning the basics and fundamentals, but how can it be possible for people who are currently employed and getting so much pressure about being constantly productive and quick and, you know, there is already so much overload? and how do we balance this.

Speaker 4 [48:55]

You're nice.

Speaker 3 [48:56]

Nice. Nice.

Speaker 2 [48:59]

Thank you.

Speaker 3 [49:02]

I think that regardless of whether you're in academia, right, or working maybe in a company on the private sector, we live in the culture of productivity, right? You have to remain yourself productive. And I think that for me, something that was key was to understand that to remain productive, studying still counts. So, like really just that change of mentality where I was still feeling productive by learning and just saying, okay, so these two hours I'm literally studying if I will be back to school because this is actually something that is fundamental for my job and my career, change of perspective. So, I'm actually really disciplined and organized. So I blocked myself times a slot to study. The stuff that I potentially are not just something that I need right away, but I would like to understand deeper on a deeper level, and it's been great. Like I reconnected with my job a lot, I have to say.

Speaker 5 [50:11]

So talking about productivity, working fast, a question about how to deal with managers who have no idea about data science and ML, but want to see results quickly. They could be also stakeholders, not only direct managers, right?

Speaker 2 [50:30]

Just go away. Run. No, I mean, actually, the sad part, whenever you go, you will meet managers and stakeholders like this, right? So the situation that you will be surrounded by people who know everything about this and who you say A and they say B and you're like united. Unfortunately, it happens only at events like this, like PyData. This is why I love being here. Like three days in a row, you meet with people who have similar wave thoughts as you are, and when you go outside, you always run into the people who don't get this or who think, yeah, LLM, it's a magic box which should produce the result, right? When I type this, ChatGPT gave me the answer, so what's the problem with your solution, right? Yeah, it takes some, you need some patience to explain why it is not like this, why it's not so easy, why the system what you can type into is not really robust and why you need more robust system, right? And here, this is going back to what I was saying earlier, if you do your job good, right, and you are kind of acknowledged as an expert, people will accept it, right? So they might disagree or they might don't like that things take time, but they will appreciate you as an expert and they will kind of say, okay, if you said this, then this is most probably what it it takes and then we will wait and then of course we have a tricky part here how to give the right estimate right because if you say it will take a week and you're underestimated and it takes like half of a year then of course your kind of reputation goes goes down but it's like bit by bit and yeah like accepting that people want things fast and it's your job to explain that it's not possible

Speaker 3 [52:19]

But if I might chime in, this is something related as well as you were saying before, that one of the skills that really matters on career is communication.

Speaker 2 [52:27]

Exactly, yeah, absolutely.

Speaker 3 [52:27]

Exactly. Yeah, so this is where strong communications skills pay off a lot. So don't underestimate that.

Speaker 1 [52:36]

Yeah, and also to start thinking about, instead of talking about the models and the work itself, just start thinking, talking more about the value and the outcome and try to get, like, because you're doing good work, so try to get yourself into being trusted that whatever you say is correct so that your manager doesn't feel like they have to verify the details of your work by micromanaging you. I mean, of course, the micromanagers out there, it's like you can't really fix them you can change them like with other micromanagers but the idea is more to to try to get yourself out of this context where you have to explain how you do your work and more talk about the results you would deliver

Speaker 3 [53:22]

Yeah, there was one course that I did early during my PhD that it was called how to get feedback, how to get your message across. 10 hours, the most useful 10 hours of my scientific life so far, I have to say. Like, because I helped me not to give feedback, but understand how good feedback was interpreted in the way that I was receiving feedback. and actually also relying on the skills that I was developing based on the feedback, the kind of feedback that I was receiving. So I was able to classify if it was a good way of receiving feedback or not. And second, the fact that, yeah, so as you were saying, focused on the values, focused on the common ground and start the conversation from there so that you actually tackle the conversation and maybe if you would like to go farther into some expertise or why you are making a decision or advocating for a decision, you can actually build trust based on the fact that you were an expert. Yeah, so a communications course is good.

Speaker 4 [54:25]

Maybe as a shortcut, the sentence, what would you like me to deprioritize for this? When they come with new ideas all the time, it's like, sure, I can take this on, but what other task do you want me to drop? Because you paid for 40 hours. I know we all work too much, but I literally just burned out, so maybe don't. And like tell your manager that them always changing your opinion, your direction has an impact and has a direct cost to the actual business as usual and to the projects that they wanted you to do two days ago. Like they can't just keep piling on to you and managing up is a skill that you have to learn unfortunately.

Speaker 1 [55:15]

And I would like to add here a proper prioritization. When you prioritize things, the most important thing is that no two things are equally of the same priority. So prioritizing really means 1, 2, 3, 4, 5. And it doesn't mean 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2. So really ranking.

Speaker 5 [55:40]

Coming close to the end of our session, there is actually a very popular question here, which would take the rest of time, I guess, to answer. What are the challenges you are still facing as a woman in building a career, and how do you deal with them?

Speaker 4 [56:00]

My challenge is that I burned out three times last year and I did not get promoted for doing extremely valuable work. So I am learning to devalue the hustle and devalue the productivity cult that we live in. And yeah, I'm actually working part-time now and I'm actually loving life right now. So consider that. Like, as someone that has always been a little bit insecure about how much I work and define my worth over my work I've always thought oh no I wouldn't be able to like clock my time because like oh sometimes I need to take breaks which everyone does they just don't tell you like getting a coffee is a break learning breaks and like breaks it's still productive it's still important so really re-evaluating first of all like the value of yourself and productivity and all of that and, yeah, kind of seeing what you can do and then maybe actually tracking your time because, like, I thought, oh, yeah, I'm not doing 40 hours. I'm probably defrauding my employer. And then I realized I'm, like, at 60 when I'm having a slow week. So, yeah, that's kind of where I'm at at the moment.

Speaker 1 [57:22]

Life hack? I mean, I always consider getting a coffee to be part of work time. Just saying. You're thinking about work on the way to coffee, it's work.

Speaker 2 [57:36]

I think my main challenge is that I started a new job just in the end of the last year, and it's a big company, and of course, I'm now in this fascinating journey of just understanding how it works, you know, and understanding, like, the cultural code and other things. I do enjoy it a lot, but of course, it's a lot of, you know, sometimes I look at my calendar and don't really understand what have I done. I was talking all day, but it looks like I did not do actually anything. I thought I did, right? So all this talking was for something and I think accepting this very new role for myself and also redefining what being productive means now and what my main value and how I can contribute to the success of my team, this is of course a big, big challenge which is also quite fascinating for me.

Speaker 3 [58:27]

Yeah, so, I mean, the academic world is tough, and I think that is overall well-known, and of course, they are, from the objective point of view, they are studies about, that basically showcase the challenges that we have to face on academia, but for me, something that I was struggling a lot was to realize, I mean, was facing the problem that, okay, I'm doing a lot, but it's never enough. Like my to-do lists never get complete. And as a person that works still with a pen and paper agenda, like it was really demotivating to see in tasks that I have to migrate and migrate the week after and the week after. I was like, wow, what am I doing? Because I'm working a lot and I don't reach the goals. And that just simply started to be a little bit more kind with myself. kind with yourself might sound like a really straightforward sentence, but it is not. Which is I started practicing a really stupid exercise, that it was instead of writing what I have left to do the week after, I started writing at the end of the week the things that I had achieved. And my first reaction was I need a break. So I think that sometimes most of the problem itself roots on the perspective and the expectations that we have on ourselves. And I think it's okay to make sure that we understand that we are humans, that we need to basically do our best and be professional, but doesn't mean that everything has to be perfect or that it has to be in the time scale that we believe that it is the one that we have to accomplish based on what productivity demands. So just make the exercise at least once a month, maybe not every week. Try to write down what you have actually achieved and rejoice with the moment. Like break, like take a coffee and say, fuck, I did it.

Speaker 4 [60:35]

Can I add a real quick tension on that one? Because doing that is actually really fantastic, like writing a work log, what you've achieved in a week or in a day, because it's also a fantastic communication tool with your manager if you have difficult managers. And for annual reviews, if you're in a system like that, especially as a PhD and stuff, you have a full log of your last year. You can throw it into ChatGPT to give you a summary and everything. It's like, I mean, we live in an age of AI, right? but it's such a valuable tool there's some really good talks and blog posts about it by Julia Evans

Speaker 5 [61:15]

That was a nice wrap. I personally want to thank you for all the insights you shared, all this frustration. I am a long-term unemployed person, and seeing your patience and Zen aura, I got inspired to take a deep breath and just be patient. And there are still some unanswered questions. I'm going to share them with you. Maybe you can answer them on Discord or LinkedIn. So follow the panelists and thank you again for your time and for engaging.