Software engineer who builds AI products end to end: the agent tooling, the backend that serves it, and the apps people use on top. I also like going one layer down, which is why I wrote my own compiled programming language.
What I'm working on
- 🧰 Hirara: a self-hosted toolkit for AI agents. 37 shipped tools (web search, fetch, PDF, OCR, Office documents, sandboxed code execution) behind one Python SDK and an MCP server, with no third-party API keys. On PyPI.
- ⚙️ LukeLang: a reactive, compiled language where one dependency graph runs from a database row to a rendered pixel. Compiles to C, then to native binaries with arena memory and no GC. Reactive semantics are pinned by 14 conformance programs in CI. lukelang.org
- 💬 ClawLink: a real-time messaging platform for AI agents. Channels, DMs, presence and read receipts over Socket.IO, with a Next.js client, an MCP server and a CLI.
- 📱 At work I design the architecture of a cross-platform AI learning product: a FastAPI backend with pgvector retrieval and streamed LLM responses, and a Flutter monorepo that ships iOS, Android and web from one codebase.
Stack I reach for
Python TypeScript C++ Dart Go · FastAPI Node.js Next.js React Flutter · PostgreSQL pgvector · Docker Nginx GitHub Actions · MCP RAG LLM tooling
Elsewhere