Department: Field Engineering

Location: United States. Preference for the San Francisco Bay Area; remote within the US considered.

The role

Turing builds the training and evaluation data the frontier AI labs use, and those labs keep asking us the same thing: run your newest model on our data and tell us what it means. In one week in late September that meant running Grok 4.7 for xAI across a set of our datasets and writing a package against Thinking Machines' model. The Field Engineer is the person who does that work and takes it to the lab. If we hand you five models and one of our datasets, you can benchmark them and write the report a client lab would read.

This is not a research role and not a forward-deployed role. You will run evals, move data through our internal systems, turn results into client material, and join sales calls as the technical person.

What you'll do

• Run evaluations and benchmarks: wire up a lab's model API across providers, run it across Turing datasets, and move data through our internal data systems.
• Turn the numbers into something the lab can use: a report on where the model falls short and which data would help, answers to the lab's technical questionnaires, and subsets of our data pulled by difficulty or by our best data points.
• Check the provenance of what we ship and keep the catalog, the website and the data hub current.
• Join sales calls as the technical person: know what data exists, in what shape, and what can be done with it, and walk the lab through that live.
• Own last-mile delivery and packaging, so what the lab receives is what it asked for.

Who we're looking for

• Data scientists, AI engineers, ML engineers and analysts who have pivoted into AI are all welcome here. A year or two at a startup where you owned something end to end counts for more than pedigree.
• Hands-on Python with model APIs and inference endpoints (OpenAI, Anthropic, Bedrock, vLLM), evaluation and judging pipelines, pandas and SQL.
• At least one piece of work where the model or the dataset was the thing you were working on: an eval you built, a benchmark you ran, a model you served or fine-tuned, a forecasting model you delivered to a stakeholder, or a dataset you owned end to end.
• Clear written and spoken communication. You will present results to researchers and sit on sales calls, and you should be comfortable doing that, not necessarily experienced at it.
• Scrappy and dependable. Junior-to-mid individual contributor, roughly one to five years after your degree; strong new grads with shipped work considered.

How we hire

A 30-minute recruiter screen, a conversation with the hiring manager or a teammate, a take-home project modelled on the work itself, and a final conversation where you present that project.

How to apply

Send a resume. If you have a model comparison, an eval you built, or a write-up where a model failed and you showed why, send that too; if you do not, we will give you a short exercise in that shape.