Remote: Yes

Willing to relocate: No

Technologies: Python, FastAPI, TypeScript, React, LangGraph, RAG, Chroma, Pinecone, pgvector, PostgreSQL, Redis Streams, Anthropic/OpenAI/Gemini APIs, MCP, LLM-as-judge evals, AWS Lambda + Bedrock, Docker, GitHub Actions

Résumé/CV: https://sjtroxel.github.io

Email: sjtroxel@protonmail.com

---

I'm basically self-taught / career-changer. Everything I have built is deployed and clickable — no private repos, nothing "available on request."

Most of my time goes into making sure the system does not confidently make something up. In Patchwork Assurance (https://patchworkassurance.com), a compliance tool for US AI-regulation law, the decision about which laws apply to a company is computed in plain code rather than by the model, so the verdict cannot be hallucinated — the model only writes the cited explanation around it. I made the multi-agent version the default only after a judged evaluation showed it beat the simpler single-pass version: groundedness 97.9% vs 95.9%, citations 100% vs 97.7%.

Same idea in the others. Heritage Odyssey (https://heritage-odyssey.vercel.app) does multi-tenant RAG over private family trees and hands back to the user when retrieval is weak instead of inventing ancestry. Wildlife Sentinel (https://wildlife-sentinel.vercel.app) runs a five-agent swarm on Redis Streams against nine government data feeds, and scored its own predictions against what actually happened afterward.

No commercial engineering experience — this is all solo work I built and operate end to end. Looking for a remote AI engineer role.