AI Engineer building measurable, local-first LLM systems.
I have around ten years of software engineering experience, from frontend development and technical leadership to building AI systems in production.
My current focus is LLM evaluation, agentic systems, deterministic quality gates, and self-hosted AI infrastructure. I care less about whether a demo looks convincing and more about whether an improvement can be measured and reproduced.
A local-first CLI and MCP server that turns meeting recordings into transcripts, summaries, action items, and decisions—without sending data to the cloud.
It is the reference implementation of the memnex v0.2 open specification and produces portable meeting outputs with a complete provenance chain.
TypeScript · Node.js · Whisper · Ollama · MCP
An observable agentic pipeline that converts photos into pixel art using local models.
A vision agent selects the rendering style, a deterministic tool creates the image, and an independent quality gate evaluates the result and triggers retries. Every decision is recorded and streamed to a live dashboard.
TypeScript · NestJS · React · Ollama · Redis · BullMQ · sharp
- LLM evaluation and golden-set design
- Agent testing and deterministic quality gates
- Mechanistic interpretability
- Local inference and self-hosted infrastructure
- Privacy-focused and local-first software
I run a dual-GPU local inference environment for LLM, embedding, vision, and speech workloads, alongside a Raspberry Pi–based self-hosted stack.
My tooling includes llama.cpp, vLLM, Ollama, Docker, CUDA, Tailscale, Prometheus, and Grafana.
Before focusing on AI engineering, I spent several years building and leading frontend systems. This gives me a product-oriented perspective on AI: models matter, but reliable architecture, observability, evaluation, and user experience matter just as much.