Spore.Bio a republié ceci
I’m happy to celebrate my first year at Spore.Bio 🎉 It’s been a crazy adventure but not always an easy one 👀 When I joined the company, everything was to be done ! We wanted a machine learning model to detect bacteria and a robust data pipeline… Except that we had close to no data at the time ! 🤷 Indeed, at Spore.Bio, we are building our own protocol to prepare biological samples 🦠 and our own optical setup 🔬 to create the data that is then analysed in data science 🤖 ! And to be honest, this is the part that I like the most about the project and the company 😄 As a data scientist, having complete freedom in the way the data is generated is a blessing because we have basically no limitations in what we sample, BUT it’s also a curse because no long term assumption on the data can be taken ! When designing our computer vision model, we used a funnel approach to stay as general as possible and keep our options open until the end. It was not always easy to forecast what would stay an invariant and we often had to refactor or rollback some features. I remember we started with a dataset of a few number of very heavy files (>1GB) for which we designed a very efficient caching system… database efficiency was not a subject at all for such few samples. But a few months later, we started working with a new optical setup producing a legion of small files ! We had to review our database architecture and refactor our data pipeline entirely 🥸 Designing our own biological and optical protocols is also a great occasion to learn more about other fields of science. I take this occasion to warmly thank my team mates Marine Pottier, PhD, Emilia Humeau, Massilia Hamdani, PhD, Marie Fournier and Maxime Mistretta, PhD. for everything that you taught me in optics and microbio 😊 If you want to solve new challenges everyday and work in a multi-disciplinary team, don’t hesitate to apply to our open positions !