Efficient deep learning KV-cache compression, quantization and LLM inference that survives contact with real hardware.
ML researcher and engineer (MSc AI, First Class Hons). By day: perception and autonomy at Syos Aerospace — real-time models on Jetson / TensorRT in production. Alongside that, research on efficient inference and contributions across the open-source ML ecosystem.
Open source device-agnostic (Apple Silicon / MPS / CPU) fixes, model support and KV-cache tooling. all open PRs →
Models & datasets K2-geoscience-7B (4-bit MLX), the first MLX build of a geoscience LLM, plus open earth-science datasets · full profile →
Stack Python · PyTorch · Transformers · MLX · TensorRT · ONNX · C++ · Docker
Connect LinkedIn · Hugging Face · ORCID
off the clock: surfing · football · somewhere on the coast