AI Engineer · contract-first · remote US

Production LLM systems, agents, and evals

I build Python and FastAPI systems around LLMs: retrieval, tool use, evaluation gates, and human handoff. Paid delivery first. Public GitHub second.

Three paid contracts. Named flagship: Acuity Real Estate, January to March 2026. SMS qualification with Spanish detection and bilingual handoff for client-reported 500+ inbound leads, with CRM integration and human handoff. Public inspection starts at DocExtract, llm-reviewer-path, and mcp-server-toolkit.

Inspect projects GitHub Resume

Recruiter proof

Three paid contracts, then public repos

Acuity is the named flagship. DocExtract is public offline-eval proof. EnterpriseHub is paid work with a replay-only demo. Then clone the three public repositories. Independent repos stay independent unless a card says paid.

Paid engagement Jan to Mar 2026

Acuity Real Estate

SMS qualification with Spanish detection and bilingual handoff for client-reported 500+ inbound leads. FastAPI, Redis, and GoHighLevel, with broker collaboration and a human handoff path when the bot should stop.

1,700+ tests at handoff; audited 226 existing GoHighLevel workflows at handoff (archive-verified). This is a named delivery story, not a public clone target. Inspectable proof sits in the public repos on the right.

Full portfolio context →

Paid · public eval

DocExtract

Measured 95.5% weighted field-level accuracy on a 28-case offline replay; separate 202-case evaluation corpus (151 golden, 51 adversarial); 80% CI coverage gate.

GitHub →

Paid · replay demo

EnterpriseHub

Confidential client, 2025 to 2026. Paid AI application work. The public surface is a replay-only demo. It does not show live client outcomes.

Replay demo →

Public proof

llm-reviewer-path

Clone, install, and run pytest with no API key. Review eval gates, action boundaries, retrieval failures, and scoping receipts.

pytest No API key
GitHub →

Supported GitHub artifact

mcp-server-toolkit

Python MCP server framework. GitHub is the supported artifact, not a PyPI snapshot. Python 3.10 to 3.14.

Python MCP
GitHub →

Where to look

Three next clicks

Portfolio

Project index

Flagship order first, then independent repositories as secondary proof.

GitHub

Public source

DocExtract, llm-reviewer-path, and mcp-server-toolkit are the first-click repos.

Resume

One-pager PDF

One-page PDF. AI Engineer. Contract-first.

Availability

Open to remote contract work

AI Engineer roles centered on production LLM systems, agents, and evals. Cathedral City, CA. Remote nationwide. Contract-first; also open to remote AI Engineer roles that match this stack. Recent paid work: Acuity, DocExtract, and EnterpriseHub.

Email caymanroden@gmail.com LinkedIn

Working stack

  • Python and FastAPI services
  • Retrieval, structured outputs, and tool use
  • Offline evaluation and adversarial fixtures
  • Human handoff when the model should stop