π I build GenAI systems that are measurable, debuggable, and resilient under real traffic.
Core
Retrieval & Data
Cloud & Deploy
I build agents by day and evaluate them by night π€π
| Project | What it does | Stack |
|---|---|---|
| mcpsum | Stops MCP servers from rug-pulling your AI agent: a lockfile + runtime reference monitor for MCP tools (like go.sum, for the tools your agent trusts) |
Rust Β· MCP |
| VibeForge | Mood-to-playlist agent: Mood Analyst β Music Curator β self-correcting Critic loop (featured in 500 AI Agents Projects) | LangGraph Β· Groq Β· FastAPI Β· Redis |
| MergeCode | Maintainer-grade review for AI-generated code: runs your checks, catches the tricks that get patches to green (deleted asserts, skipped tests, new deps, CI edits), returns MERGE / REQUEST_CHANGES / REJECT with evidence | TypeScript Β· Node.js |
| AI-Task-Architect | Turns natural-language prompts into importable n8n workflow JSON | NestJS Β· FastAPI Β· GPT-4o Β· n8n |
π€ Day job AI Engineer @ Adamson
π Side project mcpsum β lockfile + runtime monitor for MCP tools (Rust)
π Building My own trading journal with a multi-agent research desk
π¬ Ask me about MCP security, LangGraph, RAG, Multi-Agent Systems, Amazon Bedrock
β‘ Fun fact I build agents whose job is to critique other agents
π§ Principle Clean abstractions > fancy demos
Retrieve. Reason. Reflect. Repeat.
π The graph always has one more node.



