AI-native full-stack engineer with 4+ years of software engineering experience. I build product interfaces and backend workflows with TypeScript, Node.js, and Python, with public work in retrieval, citations, abstention, and deterministic LLM evaluation.
My public references make retrieval and evaluation behavior inspectable: cited source IDs, explicit no-context abstention, deterministic fixtures, typed reports, and CI gates.
Why agentic engineering changes the way software teams structure repos, feedback loops, and delivery systems.
A practical blueprint for keeping autonomous coding workflows reliable, reviewable, and safe in production codebases.
The featured work is intentionally runnable, deterministic, and explicit about where reference behavior ends and production hardening begins.
Relay Desk is a live TypeScript support operations console with cited streaming, structured actions, human approval gates, failure recovery, and clearly labeled synthetic evaluations.
LLM Reliability Lab evaluates citation validity, expected-fact coverage, abstention, latency SLOs, and estimated model cost without an API key or LLM judge.
Production RAG & Evals is a reference implementation for chunking, deterministic embeddings, lexical and vector ranking, RRF, citations, abstention, offline retrieval evaluation, and Dockerized delivery.