I build local-first tools for agentic AI coding, AI-assisted workflows, and self-hosted model infrastructure.

Currently working on:

• WorkStream - a self-hosted system for running AI workers through structured coding flows, isolated git worktrees, review steps, and PR handoff.
• LM Bridge - a personal AI chat platform connected to LM Studio, with memory, tools, RAG, Telegram/web interfaces, scheduled jobs, and sub-agent delegation.
• EndStream - an original web strategy/card game with a TypeScript engine, React client, persistent server state, and multiplayer/LAN workflows.

I'm interested in the practical edge of AI coding: agent loops that actually finish work, tool execution that can be trusted, local models, MCP, memory/RAG systems, and human-in-the-loop software development.

Main stack: TypeScript, React, Node, Express, Supabase/Postgres, SQLite, Vite, Playwright, LM Studio, Claude Code, MCP.