AI Engineer building production ML systems — computer vision pipelines, LLM/RAG applications, and the MLOps to run them. I care about the unglamorous parts: model licensing, latency budgets, and what it costs to run at scale.
- Computer Vision — object detection, face recognition, and tracking pipelines; picking models for accuracy and license/latency/cost constraints, not just leaderboard rank
- LLMs & VLMs — RAG pipelines, agentic workflows, MCP integrations, and knowing when a vision-language model earns its compute
- MLOps — model serving, containerized deployment, monitoring, and CI for ML systems
- Full-stack when it ships the model — SvelteKit/React frontends, Cloudflare Workers, Postgres/Supabase
| Project | Description | Tech |
|---|---|---|
| Junto | Keyboard-first task tracker with a built-in MCP server — manage tasks from Claude. FTS + pgvector semantic search, realtime sync, PWA |    |
| mcp-worker-template | MCP server template for Cloudflare Workers with zero runtime dependencies — hand-rolled Streamable HTTP transport in ~190 lines |  |
| cv-model-licenses | License-aware registry of CV models — which detectors, face models, and VLMs you can actually ship in closed-source commercial products |  |
| PlugPulse | Community-verified EV charger reliability — a freshness-weighted trust layer on Open Charge Map data |  |
| Smart_Minds | AI recruitment platform — resume parsing, LLM candidate-job matching, RAG chat over a talent pool |   |
| PodPilot | AI-powered podcast platform with text-to-speech generation and discovery |  |