AI / LLM Engineer — I build production agents, RAG pipelines, and the tooling that keeps them reliable.
Software Engineer at Enidus USA, where I build enterprise AI for T-Mobile for Business — multi-tenant LLM tool-calling backends, hybrid retrieval, and LangGraph agents. M.S. in Computer Science from Indiana University Bloomington.
| Project | What it does |
|---|---|
| tracelint | A deterministic linter for agent runs — reads the execution trace and fails CI on structural bugs (ignored tool errors, schema violations, loops, duplicate side effects). No LLM judge. |
| muteval | Mutation testing for LLM eval suites — degrades a system on purpose to check your evals actually catch regressions. Refuses to score on a red baseline. |
| Retrieval-Lab | Benchmark RAG retrieval configs on your own corpus, with deterministic, stage-level attribution for every failed query. |
| goldset | Turns production traffic into a versioned, deduplicated, coverage-tracked LLM regression suite — with a mandatory human-confirmed golden gate. |
| smolAmem | Multi-tier long-term memory for agents: working, episodic, and semantic. Published on PyPI. |
| ToolPicker | Hybrid BM25 + semantic router for agents with too many tools for the context window. Published on PyPI. |
| DocChat | Answers library questions from the exact version your lockfile pins. VS Code extension + MCP server. |