From finance to ML and Computer Vision

Hello, I’m Guillaume. I studied mathematics in the early 2000s and worked in finance for 12+ years. Then I needed a change and decided to go back to my roots and embrace the inner geek that was always in me :)

So I went back to school and got an MSc in Machine Learning (Telecom Paris). This was super fun, I enjoyed so much learning new things in a field that is both practical and theoretical.

I then had the opportunity to work at an amazing healthcare start-up where we applied Deep Learning algorithms on time-lapse microscopy. I met wonderful people there and my passion for Computer Vision grew stronger. We were lucky enough to publish a few papers in the process.

Nowadays I work with satellite imagery, still developping deep-learning models and still having a lot of fun !

Computer Science

I enjoy improving my programming skills in every aspects: performance, design, testing, CI/CD… I mainly code in python and Rust. I developed a couple of libraries such as video_reader-rs and perfect-matching to solve some pain points I encountered on several projects.

• video_reader-rs is a fast video decoding library written in Rust with python bindings, written as a replacement to the venerable decord library.
• perfect-matching is a Rust library which offers several Linear Sum Assignements algorithms. I wrote it as an exercise to get a better grasp on this family of algorithms, see if I could improve performance compared to scipy (and others) implementations and play with SIMD instructions.