How we built SEER — the evaluation rig that lets us change Scout’s prompts and swap the models underneath it without breaking its judgment, and eventually taught an agent to improve its own prompts overnight.

Scout vets workers — but vetting is just one step in a much longer funnel. A new series on the agent infrastructure Traba is building to optimize the entire industrial worker lifecycle, and what “Physical AI” really means when your decisions affect real people's livelihoods.

Traba is reimagining industrial staffing by building Scout - an AI recruiter that conducts thousands of real-time interviews in parallel, qualifying workers with the speed and consistency of automation while preserving the depth and personalization of human-led conversations.

Traba transitioned from Firestore to PostgreSQL to manage escalating costs and improve scalability. The migration reduced database expenses by over 70% and enabled greater control and efficiency by switching to a fixed hourly billing model. Our journey highlights key insights for optimizing database architecture for growth.