Open to lead and principal AI engineering roles Fort Worth, TX — local

Let's talk.

If you are building LLM systems that have to hold up in production, or the evaluation and platform work underneath them, I would like to hear about it. Email is the fastest way to reach me.

What I'm looking for

Role
Lead or Principal — production LLM systems, evaluation and reliability, and the distributed platforms they run on
Work model
Remote (US), or hybrid in Dallas–Fort Worth
Based
Fort Worth, TX · US Central
Overlap
A full workday with US Eastern and Pacific; mornings with UTC+0 to +2
Reply window
One business day

Good reasons to email me

  • You have an LLM feature that works in a demo and you need it to survive regulated production: data classification, model risk, an on-call rotation.
  • You cannot tell whether a model, prompt or schema change made your system better or worse, and you want the harness that answers it.
  • You want to talk about agent tooling and MCP, or about anything in boost.

What happens after you email

  1. You send anything from one line to a full role brief. No form, no template, no screening questionnaire.
  2. I reply within one business day. Usually with a direct answer rather than a request to schedule a call.
  3. A 20-minute call if there is a fit. I will have read the job description and whatever your team has published.

Questions people ask

What have you actually run in production, versus explored?

In production: the document-intelligence pipeline and its evaluations, MCP integrations across Jira, Confluence and GitHub, and eleven years of distributed systems behind them.

Not production: no vector database, and no LangChain or LangGraph beyond a hackathon prototype. The guides in the Library label reference designs as reference designs, and Cerebro is an independent research project, not employer work.

Can I see code?

boost is open source: a from-scratch BM25 retrieval engine in standard-library Python, an MCP server, and a CI gate that scores retrieval on recall@k, MRR and nDCG@k against a graded golden set. The committed baseline is at evals/baseline.json.

Are you open to contract or advisory work?

My focus is a full-time lead or principal role. Ask anyway if the problem is interesting — the worst outcome is that I point you at someone better suited.

Do you need visa sponsorship?

No. I am authorised to work in the United States without sponsorship.

Not ready to email?

Reasonable. Here is the work itself — each of these answers a different question about how I build.

Library Technical guides

How I explain a system to the engineers who have to run it. Start with the pgvector guide, written twice: plain English and engineering.

 Open source boost

A CLI and MCP server whose retrieval engine is gated in CI on recall@k, MRR and nDCG@k. The evidence that I ship evaluation, not just talk about it.

 Independent project Cerebro

Promotion gates for an autonomous trading system. No strategy reaches capital until it clears four of them, including a human sign-off.

Either way, thanks for reading this far. — Jonathan