Yomi Lebi AI Engineer — Contract Tucson, AZ — Available now

Yomi Lebi

AI engineer for contract work. I build LLM agents and AI tools, and I measure the results.

(01)

Selected work

(02) About

I make LLM systems show their work: tested offline, scored against labelled data and costed per run.

I studied computer science at the University of Arizona, then left in 2026 to work on AI engineering full time. Since then I've designed and shipped LLM tools end to end, from design doc to tested, measured release.

Languages
TypeScript, Python, SQL
Web
React, Node.js
AI
Evals, LLMs, agents, LangGraph, LangSmith
Testing
pytest, ruff, mypy, GitHub Actions

(03) How I work

I design, decide and verify. Claude Code writes the code.

I set the architecture and work the details through with Claude, which drafts the design doc for me to revise. Claude Code then writes the code. That split makes me quick: I go from design doc to a tested release in days.

The decisions stay with me. I turn down designs that give the model more control than the job needs, and keep the rules in code: code lists the legal moves and the model picks one.

I check the work three ways: offline tests that run on every change, hand-labelled data with a held-out set, and live runs I watch myself.

Design
Mine, drafted with Claude
Code
Claude Code
Checks
Tests, held-out labels, live runs