Women in Tech: Talking with Kasia Kulma about her journey as a Data Scientist

Reading Time: 6 minutes

Women in Tech: Talking with Kasia Kulma about her journey as a Data Scientist

Women in Tech Thursdays: Talking with Kasia Kulma about her journey as a Data Scientist.

Kasia Kulma is a Senior Data Scientist with a specialisation in R and PhD in Ecology and Evolutionary Biology. She has industry experience in pharma, insurance, finance, aviation, food delivery, environmental services and in her spare time contributes to opensource around Rstats and climate. In addition to that, she’s also a mentor and organiser aR-Ladies London meetup and blogger at https://r-tastic.co.uk/.


Tell me about your journey in the world of tech. How did you start and what was your motivation, what attracted you to the tech career you are in? 

First of all,  I want to say that I never intended to work in tech! All my life I was fascinated with biology – by the time I was 15 I knew what exact field I wanted to be in, what university I’d go to, what research team I’d do my Master thesis, etc. I executed my plan to the dot and obtained my PhD in evolutionary biology from the Uppsala University in Sweden in 2013. Something that I didn’t take into account was how challenging finding a long-term contract can be in academia. My son was just born (2 weeks after defending my PhD thesis!) and we were about to move to the UK. My partner was also an academic and I thought that one researcher in a family is more than enough so I decided to leave science. But leave to do what? I had to think hard about what I enjoyed doing most and what transferable skills I could offer. My PhD was filled with analysing messy data and performing statistical analysis and it was by far the best part of my studies (except spending 2 months a year in the woods doing the fieldwork, that is!). By then I had programmed in R for several years and I was keen to learn about this mysterious Big Data that everyone was talking about! The biggest appeal to do that was an intellectual challenge and the constant learning that data science requires, just like in academia. However, unlike academia, data science expects quick iterations and tangible impact of your work, even if the solutions are not perfect and I liked that approach much more. 

 

Do you think that tech is biased towards one gender over the other and if so, in what ways have you experienced it?

Of course, tech is biased! Some sources report as little as 16% of the IT workforce being women [1]. But after more than 7 years in the field, I’m still trying to understand the full scope of factors that affect these numbers. There’s no doubt that lack of role models plays a huge role – it definitely did for me. When I first started to learn programming in my early twenties, guess what, I wasn’t very good at it. But in my mind, this was expected (“Coding is for guys”) and I was almost annoyed that I had to learn something that “I was not made for”. My PhD supervisor was a woman and a great mentor but she never picked up coding herself. It was only after I’ve joined R-ladies that I realised that there are other women learning – and enjoying! – coding and that was truly transformative to me. 

Societal expectations about women tech skills are something I think about often. As a data science consultant, I often am the only woman in a client meeting or a dev team. I have to confess that every single time I try extra hard to convey that I’m a capable and hands-on developer and not “just” a good communicator or a project manager. Once, one of our salespeople wanted to recommend me to a client by saying “Kasia is one of our most empathetic consultants”. Comments like that are meant well but I think they do women in tech disservice, as they only reinforce the notion that the soft skills are something that makes them stand out, not their technical capabilities. Having said that, the indsutry is changing and there are more and more places that are trying to bust these myths.

 

What, in your opinion, might make other women hesitant about pursuing a tech career? There are so many Women in Tech initiatives right now (and here we are, adding yet another one) – and yet we are still working to close the diversity gap. What do you think is holding women back?

The two biases I mentioned above – lack of role models and the perception that the hard tech skills are not women’s strength – are very important, in my opinion. And this creates a vicious circle common in all environments that lack diversity – whatever practices are exercised or language is spoken, they are considered normal and expected from everyone else. It also creates a confirmation bias in the recruitment process: if you don’t speak that language or show these behaviours, you don’t fit in. A good example here is job specs that are often very specific and hyperbolic in regards to required experience or technology stack that a candidate should know. It’s a passing joke that if a man fulfils only 50% of the requirements he’ll still apply for the position, but a woman will do the same only if she feels she fulfils 100%. More diversity-friendly job specs focus on broader skills and qualifications, embrace candidates from various backgrounds (this includes the programming languages they know) and will emphasise that if the role sounds interesting, you should apply even if you don’t match all the criteria. That’s how you can learn a lot about a company even before the first job interview. 

On that note, I want to add: women hold themselves back, too. It’s ok not to know everything. It’s fine to not always ask a super clever question or or simply voice your opinion even if no one explicitly asked for it. And yes it’s ok to apply for a job that does not match your experience 100%! At the end of the day, you want a role where you can grow and learn, not just apply skills that you already have. Even if you have to learn a lot, you can still bring a lot of value to your project, team, company. 

  

You’re a data scientist and a blogger with massive knowledge and experience and a specialism in R. But for someone who’s just starting their journey, the prospect of learning so many technologies might be a bit daunting. Have you ever felt overwhelmed or challenged yourself? If so, how did you cope with that?

Goodness gracious, of course! I feel challenged and overwhelmed most days. Let’s face it – there’s no end to what you can learn and there will always be someone who knows more about a certain subject than you. That’s just part of the game. But don’t forget that there are a lot (yes, A LOT) of people who can learn from you and your unique combination of experience and expertise. Still, I appreciate that if you’ve just set off on the data science journey you may be confused by all the buzzwords. So here are a few pointers that I’d like to share:

  • Identify what you want to learn and why: is it something that you could use on the project? Are you genuinely curious about the topic? Or do you want to learn it because you *think* that you should?
  • Stay focused: prioritize things you want to learn and break them down into smaller, more digestible chunks. 
  • Stay real: online courses are good but they can give you a false sense of reassurance that you know how to do something just because you have a certificate for it. Go and practice your newly gained skills on real examples (e.g. by using open data).
  • Stay connected: reach out to your community and don’t be afraid to ask questions. And don’t limit yourself to stack overflow, send a post on Twitter or even current or ex-coworkers if that’s going to bring you closer to the answer you need.  

Personally, I draw a line between general data science skills and those specific to a project or industry. For example, basic data wrangling, version control and software engineering principles (i.e. how to modularise, test and document your code) are needed on *any* data science project. Those are the skills I’d prioritize in any learning plan. 

On your blog, you write about the exploration of data using R. What is your favourite example?

Can I pick more than one (laugh)? By far, the most popular post on my blog is “Automated and Unmysterious Machine Learning in Cancer Detection” – an article that highlights the importance of making your model prediction human-interpretable to better understand, evaluate and improve your model. In this case, I used a framework called LIME (Local Interpretable Model Explanations) for interpreting predictions of breast cancer generated by a neural network. But I’d encourage everyone to be creative in ways you can play with your data! For example, here I compared word frequencies between Star Trek and Star Wars movies. Or here I explored London crime data using heat maps. Or here I built a text classifier distinguishing between Donald Trump and Hilary Clinton’s speeches. I had so much fun working on these side projects and all of them relied on open data and opensource technology.  

 

What advice would you give to your younger self?

Don’t worry too much about not knowing everything – no one does. 

You’re doing better than you think you do, so don’t let self-doubt hold you back. 

Don’t let labels define you or put you off doing what you want to do.

And maybe one more: it’s ok to change plans sometimes, ha! 

You can follow Kasia on Linkedin and access her great blog here.

Thank you, Kasia!

[ninja_form id=4]