"Computing machines will eventually understand better than humans, but feeling belongs solely to humanity."
- Currently: M1 Informatique (Parcours ICO) – Université de Montpellier 🇫🇷
- Background: Studied M1 Electronics (EEA), providing a strong cross-disciplinary foundation in hardware design.
- Mission: Making generative AI accessible to everyone through open-source tools.
- Contributions: Developing the ComfyUI ecosystem, contributing to llama.cpp, building numerous personal open-source projects, and creating hardware generation frameworks.
SpinalML is an open-source framework for generating AI hardware accelerators on FPGA (based on SpinalHDL).
- Hardware Demo: Real-time inference (MNIST W4A8) on a Tang Primer 20K board with a web interface.
📅 View the development roadmap (August 2026)
- Aug 3: Project kick-off.
- Aug 5: Initialization of the first self-hosted CI.
- Aug 10: Integration of the Python verification pipeline.
- Aug 17: Addition of formal verification (SymbiYosys).
- Aug 19: Implementation of the Attention operation.
- Aug 24: First MNIST BF16 test successfully validated with Verilator.
- Aug 30: Milestone: Functional hardware inference of MNIST in W4A8 precision.
- QTRobot-Interaction - Human-robot interaction pipeline.
- QT-Jacket - Machine Learning embedded on the QT Jacket.
