Public repositories

A multi-domain agent-evaluation engine — one plan-act-observe loop over real tools, mechanical (keyless) scoring, deterministic replay, and pluggable domain packs (generic, industrial, trust & safety).

An MCP server over a governed dbt + MetricFlow semantic layer (DuckDB) — three tools and nine metrics, with lineage back to the dbt nodes and no SQL composed from model or user input.

Reproducible llama.cpp benchmark harness and raw results for the AMD gfx1150 iGPU — a roofline, an 8-model ROCm-vs-Vulkan matrix, and a 2.43× speculative-decoding result.

A SvelteKit metrics console that is itself an MCP client of dbt-semantic-mcp — a catalog, query builder, lineage view, and hand-written SVG charts; every number sourced from the governed layer.

An OKF-compatible knowledge-context format with a bidirectional adapter — reads memory, skills, and vault surfaces into a unified format and exports losslessly; built to drop in as a hot-swappable Spaces pack.

Agricultural CAN/ISOBUS protocol library — a DBC corpus mapping raw frames to named signals, scoped to pure-standard PGNs only (ISO 11783 + the J1939 ag subset). Scope is enforced by CI, not review.

A single-file, no-build renderer that compiles LLM-authored specs into self-contained HTML — forms and dashboards from one engine; ships a dependency-free MCP Apps server.

Hold-to-talk local dictation for Wayland on GNOME — CPU-only faster-whisper, audio kept in an in-memory buffer, no network after the one-time model fetch, user-scope only.