#42 Paris Women in Machine Learning & Data Science: full web apps with open source, NLP for proteins, and the Bechdel AI project
What a meetup! On March 16th, despite the strike, we had our biggest meetup ever — with over 150 attendees, exploding our last record. We are very proud of this, and we take this as a proof of how welcoming our community is. Thank you everyone for contributing!
Thankfully, our host, Google, put a neat process in place to welcome our attendees, leading to a smooth welcome for everyone despite the crowd.
This meetup was special for the organization team, as we welcomed two new team members: Jihane Bennis, Lead Data Scientist at Bouygues Telecom & Juliette Bassnagel, Data Scientist at LiveRamp.
This is big news for us, and we are happy to have these incredible women onboard, to help organizing the next meetups, events, and side-projects.
Back to the programme.
Despite supporting open source tools is our hearts, we rarely have talks about open source. In fact, our last meetup about open source was a year ago, when we hosted a scikit-learn sprint.
It was time to highlight another open source tool, and thanks to Marine Gosselin, it has been done. She presented Taipy, an open-source Python application builder.
Data Scientists are not Software Engineers. The data scientists can’t productionnize their work themselves, and dealing with several teams on the same project can be tricky. This might explain partly the “pilot syndrome”: the fact that most data science projects remain pilots and do not make it to the production stage, even when they proved to be working.
Marine showed us how, without a lot of knowledge in apps, it is possible to make one. First, the showed how to create an interactive and powerful interface in a few lines of code; second, how to manage your models and pipelines and create scenarios the smart way.
If you want to become a contributor of Taipy but are not sure how yet, they have mentorship program. If you are new to data science, it can be a great way to start your data portfolio.
If you want to know more, check out her slides below.
Recently, Large Language Models (LLMs) have become very popular. You have surely heard about ChatGPT or BERT — they both are LLMs.
Amélie Héliou explained how they works. LLMs are based on a way to organize neural networks, called Transformers, invented not so long ago, but already widely used due to their impressive performance. One reason is they perform well because they are able to “remember” lots of words.
How can these models be useful? Amélie introduced us to the world of proteins: in uniprot, researchers register millions of proteins. The name of these proteins are usually given by experts or rule-based annotations, and they give an idea about the function of the protein, or its composition. But they don’t have time to name them all! Here is where Amélie’s work enter the show: based on LLM, she and her team suggest a name for protein, easing the researchers’ work.
If you want to know more, check out her slides below.
Our last task was about the project Bechdel.ai, a data-for-good initiative to measure representation of women in films and media.
Have you heard about the famous, or rather infamous, Bechdel test? It tests representation of women according to three simple rules:
(1) the film features at least two women;
(2) these women talk to each other;
(3) they talk about something other than a man.
You would be surprised to know that even today, only around 50% films pass this test.
Sandrine Henry walked us through the BechdelAI project to automate the assessment whether a movie passes the test or not. It turns out to be pretty useful: as of today, we don’t have so many movies scored, as it’s very time consuming to check it.
If you want to know more, check out her slides below.
As for this and other meetups of ours, we certainly pass the Bechdel test with flying colours!
We are cooking a least one more meetup for you before the summer break. Let’s keep in touch!
If you do not want to miss our events, you can:
🔗 follow us on Twitter, Meetup, and LinkedIn
📑 check our Google spreadsheet if you want to speak 📣, host 💙, help 🌠
📍join our Slack channel for information, discussions, and opportunities
📩 send an email to the Paris WiMLDS team to paris@wimlds.org
🎬 subscribe to our WiMLDS Paris Youtube channel
📸 follow the global WiMLDS on Instagram, LinkedIn, and Facebook
🔥 share your company or lab’s job offers for free on the global WiMLDS’ website.
