Senior engineer and AI systems builder. 20+ years in production systems (banking, infrastructure). Building open-source AI tooling in evenings and weekends.

Current focus: local LLM inference on AMD GPU, language geometry experiments, code intelligence for grounded LLM workflows.

Published work:

Stack: Python, Rust, ROCm/HIP, SQLite, GGUF.

Results here are real runs on real hardware. No cherry-picking, no synthetic benchmarks. If something didn’t work, that’s in the post too.