I build AI systems, developer tools, and useful products—usually from the question: why am I still doing this manually?
My work ranges from orchestrating fleets of coding agents to distilling reasoning models on Apple Silicon. I like projects with sharp interfaces, measurable outcomes, and enough technical depth to teach me something new.
current_mode = "shipping"
interests = ["agent orchestration", "local AI", "developer tools", "product engineering"]
philosophy = "make it work → understand why → make it excellent"
|
An AI agent orchestration platform for spawning, coordinating, and reviewing fleets of Claude Code, Codex, OpenCode, and Hermes agents.
|
⌁ SQL-R1Distilled DeepSeek-V4-Pro into Qwen2.5-Coder-7B on Apple Silicon—lifting BIRD text-to-SQL pass@1 from 47.5% to 55% for $5.83.
|
⌖ SeatlineA local Cineplex seat-availability watcher that finds cancellations in the part of the auditorium you actually want.
|
A batteries-included Claude Code framework with 213 production skills, 37 specialist agents, and 95 namespaced workflows.
|
- Building local-first interfaces for multi-agent engineering
- Experimenting with model distillation, LoRA, and rigorous evals
- Turning tiny real-world annoyances into focused software
- Exploring what happens when AI tools become collaborators, not autocomplete
Less ceremony. More building.