Software engineer who designs systems to solve problems I actually have, then open-sources the interesting parts. Part-time scrum master for a fleet of AI agents that do what I tell them — most of the time.
- Forge — Project infrastructure managed as code. GitHub Projects board driven by a declarative reconciler over the GraphQL API — local manifest diffs against live state and applies changes idempotently. Three interconnected MCP servers handling data ingestion, project tracking, and structured content management.
- Agentic Workflows — Multi-agent orchestration framework using MCP (Model Context Protocol) servers as the integration layer. AI agents coordinate across services with structured tool interfaces, automatic context management, and human-in-the-loop governance. Currently running Claude, Qwen, and Gemini agents across two machines.
- personal-history-db — A personal data platform ingesting 10+ years of digital history across 28 source formats into a unified store. Plugin-based architecture with semantic search, typed-graph knowledge layer, and scheduled compute pipelines. Python, SQLite, sqlite-vec.
- WreckSys — Sequential recommendation engine for books, built on a GRU4Rec two-tower retrieval model in TensorFlow. Trained on 55.4M interaction records across 726K users. Sub-60ms API, deployed on AWS ECS.
Python, JavaScript, SQL, Git. TensorFlow, PostgreSQL, SQLite, FastMCP, AWS (ECS, ECR), Docker, GitHub Actions CI/CD. Homelab on TrueNAS SCALE — 12 services behind nginx reverse proxy.
BS Computer Science (WGU). Currently at Amazon. Spent 15 years in operations building teams and developing people — now I build applications and train models instead.