Systems & AI engineer, from CPU cycles to product
High-performance architecture consultant
South Brittany
I build herbert-rs, a local LLM inference engine in Rust and hand-written assembly, optimized at the instruction level. Faster than llama.cpp in CPU decode.