AI Systems Architect
Independent researcher building the tooling layer between language models and real-world infrastructure. Former AI Director. Guest lecturer at Harvard, MIT, Princeton, Cornell Tech, NYU, and Oxford.
Persistent memory system using PostgreSQL + pgvector. Gives Claude long-term recall across sessions with browser extension, dashboard, and cross-device sync.
6-agent tmux coordination system with 33 team templates and 19 MCP tools. Model-agnostic — works with Claude, GPT, Ollama, and local models.
Meta-MCP server that auto-generates standard-compliant MCP servers from a single schema definition. Produces tested, deployable servers on demand.
Open standard for consistent, safe MCP tool interfaces. 25 mandatory tools, modular architecture, 238 automated tests.
Production MCP server with 111 tools across 7 networks. Demonstrates MCP patterns for complex multi-system orchestration.
Single MCP interface to 106+ financial data APIs with real-time cross-platform arbitrage detection.
45 technical posts documenting the full journey from experimental MCP development to published standard and multi-agent systems.
Materials from guest lectures at Harvard, MIT, Princeton, Cornell Tech, NYU, Oxford, and Web3 Summit Amsterdam keynote.
I'm exploring opportunities at AI research labs focused on agent infrastructure, tool use, and human-AI collaboration.
Get in Touch Read the Blog