AI Systems Engineer. I design LLM infrastructure, agent systems, and production SaaS, and I ship the open-source tooling behind them.
- LLM evaluation and cost. Automated prompt regression and cross-provider cost analysis across Claude, GPT, and Gemini. (TunePrompt)
- Agent orchestration. MCP-native multi-agent systems built from reusable skills and workflows, running across Claude CLI, Gemini CLI, and Cline. (Genesis Forge)
- Context infrastructure. A self-hosted, JWT-authenticated protocol for delivering context to AI tools. (OpenContext)
- Production delivery. End-to-end services on NestJS, FastAPI, and Next.js, deployed to Cloud Run and AWS Lambda.
Full Stack & AI Engineer at Girl Power Talk, building a full-stack legal SaaS for expert witness workflows. Open to relocation.
Merged or submitted fixes in production open-source projects, focused on auth correctness, filesystem security controls, and data integrity in evaluation pipelines.
| Project | PR | Change |
|---|---|---|
| openClaw | #52513 | Fixed async gateway auth false negatives caused by SecretRef resolution blocking valid requests |
| openClaw | #52921 | Added session auto-bootstrap, removing cold-start auth failures |
| modelcontextprotocol/servers | #3678 | Added a configurable --follow-symlinks depth limit to the filesystem server |
| microsoft/promptflow | #4100 | Fixed UTF-8-BOM handling in JSONL that silently corrupted evaluation data |
| Area | Tools |
|---|---|
| Languages | TypeScript · Python · Java |
| Backend | Node.js · NestJS · FastAPI · Spring Boot · Socket.IO |
| Frontend | React · Next.js 15 |
| AI systems | Claude API · MCP server and client development · LangChain · LangGraph · Multi-agent orchestration · Pinecone · Automated prompt evaluation · Cross-provider cost analysis |
| Data | PostgreSQL · MongoDB · Redis |
| Infrastructure | Docker · Cloud Run · AWS Lambda |
- MCP Specialist, Anthropic (2026)
- AWS Certified Solutions Architect (2025)