CS engineering student at Polytech Tours and AI research intern at HEC Montréal (Summer 2026). I focus on native systems, edge AI runtimes, and low-level performance.
I build software that runs directly on constrained hardware:
- llmedge – Android wrapper for llama.cpp, stable-diffusion.cpp, whisper.cpp, and ONNX. Handles native runtime pooling and OpenCL/Vulkan/CPU fallback logic to prevent crashes on device-specific drivers.
- LightDiffusion-Next – Stable Diffusion pipeline and desktop GUI built in PyTorch. Uses custom batching and memory layout optimization to squeeze inference speed out of low-VRAM GPUs.
- EasyReader – Manga and PDF reader for Android. Integrates
llmedgeto run chapter summaries and text-to-speech entirely offline.
- Pilotrs – A
no_stdRust flight simulator and autopilot. Implements MEKF/complementary filters for state estimation and PID/LQR loops for attitude control. - FrameLimiter – macOS frame pacer for Metal games. Hooks
CAMetalLayer'snextDrawablevia DYLD injection to apply backpressure, reducing GPU power draw and heat. - CChess – A chess engine written in C using bitboards, alpha-beta pruning, and iterative deepening.
- Languages: C/C++, Rust, Python, Kotlin, Objective-C
- Domains: On-device ML, sensor fusion, graphics (Metal), system hooks, control loops
- Tools: NixOS (Flakes), Git, ADB, PyTorch, Android NDK