AI Solutions Architect

• Own the scoring rules applied to every observed answer, and keep them written down on the methodology page.
• Design how each measurement run is recorded so every citation traces back to the answer it came from.
• Turn research requirements into a measurement design the team can repeat on the next volume.
• Prepare the datasets and analysis deposited with each published research volume.
• Decide when a result is solid enough to publish and when a limitation has to be stated beside it.
• Review client measurement before findings reach the client.

• Five or more years of experience building production systems.
• Working depth in Python, system design and data pipelines.
• Hands-on work evaluating large language model output.
• A piece of analysis or evaluation you shipped that someone else could rerun.
• Eligibility to work in the United States without visa sponsorship.

• Remote only, hiring in the United States. Relocation is allowed.
• The team works from Bloomington, Chicago, San Francisco and New York.
• The team sets its own hours across US time zones, and work is judged on output.

• Published figures can be recomputed by anyone from the deposited data and analysis code.
• The same question and engine pair scores the same way under the published rules on every run.
• Client findings rest on measurement that holds up when the client checks the receipts.

The method behind these measures is published on the methodology page.

$140k to $190k a year, full time, no equity. These are the bands as posted on Wellfound.

The role is also listed on Wellfound.

What you send is kept for twelve months and used only to assess your application. See the privacy notice.