🗓️ 2024 - Present 📍 Paris - manager data scientist @ capgemini invent
🗓️ 2021 - 2024 📍 Paris - data scientist @ quantmetry
🗓️ 2020 - 2021 📍 Malaysia - deployment & research engineer @ serimax (vallourec)
🗓️ 2015 - 2029 📍 Paris - team lead for development of car breaking system @ bosch
🗓️ 2020 📍 Paris - M.S., Data Science - yotta (bootcamp data science)
🗓️ 2015 📍 Toulouse - M.S., Engineering - icam
🗓️ 2015 📍 Toulouse - R.M.S., Mechanical - university III toulouse
coding (well it seems obvious) in python, bash, SQL and typescript.
rag & agentic system with strong focus on langchain/graph/fuse libraries.
conventionnal nlp like before the invasion of the external editors apis providing llms too big to be self hosted. Yep, pepperidge farms remembers good old days when having 100M parameters for your model was huge.
code versionning (I won't pretend that I know other VCS than git, but hey, i know the difference between merge & rebase).
ro with pyomo python library.
cloud provider skills mostly on azure & databricks.
how to fine tune a mistral model on colab{:target="_blank" rel="noopener"}
- cool project where I managed to fine tune an open source mistral base model with 7b parameters , on google colab free tier.
- using quantization techniques and lora (during the hype peak) to fit all that memory on the 15gb colab GPU.
how to logging like a pro in python{:target="_blank" rel="noopener"}
- wip but the idea is to get the best out of logging in python
running (proud 2h33' marathon runner)
guitar (trying to shred as fast as possible) & harmonica (please mention Neil Young, not Bob Dylan)
video editing (premiere pro for the win), drone flying & photography
mountaineering & hiking (GR20, Tour du Mont Blanc, Rinjani ascent - & descent obviously)
horror movies enthusiast
- for specific question on my dev knowledge : feel free to open a pull request on this repo
- for others questions : my linkedin is here{:target="_blank" rel="noopener"}
About myself:
born during the early 90's, I had to be 33 to do my first porfolio.better late than never.\
This portfolio is guaranteed to be generated without any llm.