AI-native engineer in Seattle. I build MCP servers, agent tooling, and test harnesses for systems that are hard to test.

Everything below is public and checkable: test counts, CI runs, registry entries, and commits in other people's repositories.

vercel-deployment-mcp Stateless stdio MCP server, four read-only tools for observing Vercel deployments. 66 tests, CI green, MIT. Published on npm as @addiplus/vercel-deployment-mcp and listed in the MCP registry as io.github.addiplus/vercel-deployment-mcp.

llamacpp-batch-invariance Batch-invariance verifier for llama.cpp continuous batching. Per-cell diffing, a five-verdict output contract, and signed GREEN or RED certificates. Mock-sourced passes are non-promotable by construction, so the gate cannot rubber-stamp itself. 138 tests across 8 files, CI green, Apache 2.0. Ships a real Phi-4 RED divergence certificate so you can watch the gate bite.

registry-radar Next.js 16 App Router app for tracking npm and MCP registry packages: download sparklines, server-paginated version history, a watchlist, markdown export, and explicit failure states including a rate-limit banner with reset countdown and stale-while-error handling. 142 tests across 16 files, CI green, MIT. Live at https://registry-radar.vercel.app

mcp-cookie-auth-reference Reference MCP server for cookie-session systems that expose no public API, with a Hono mock target and fixtures so the suite runs offline. 24 tests, MIT. Writeup: How I built MCP servers for systems that have no API

• BoundaryML/baml#4070, escape Python reserved words in generated SDK identifiers. Open. +3,592 / -140 across 13 files, covering every identifier site in the Python generator plus a bridge fix, preserving BAML wire identity while changing host-language spelling.
• vercel/ai#15568, resolve npx and uv ENOENT on Windows via shell spawn. Open. +95 / -9, with two platform-mocked tests.
• mbachaud/Cymatix-Context#190, a Grace Blackwell aarch64 launch-blocking shim, shipped default off so x86_64 does not pay the sync tax. Merged 2026-06-09. I am one of two contributors to that project.

• prd-engine, multi-agent PRD pipeline, 10 agents and 4 intensity tiers.
• safe-rm-guard, a Claude Code hook that blocks recursive deletes, including the false-positive case the obvious regex gets wrong.
• claude-skill-pack, four Bun-native Claude Code skills.
• mcp-builder, notes on dual-transport MCP servers and the OAuth 2.1 ceremony.

Seattle. Open to AI-native engineering work.