Turing is contracting logistics and fleet-operations practitioners to ground LLM-built supply chain tools — route planners, work order managers, vendor trackers — in real dispatch logic. You author the constraints and guardrails, validate context files like dispatch logs and driver rosters, and sign off on whether AI output would survive contact with an actual dispatch desk.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work actually involves

This is not annotation of loose text. You are specifying and then auditing operational tooling that a model builds. Day to day that means defining discovery parameters — fleet size, service area, time windows, vehicle capacity — and writing the guardrails that sit around them: hours-of-service compliance, weight and volume limits, delivery SLA definitions. You then work through what the AI produces and check whether its routing logic, assignment rules, and cost formulas hold up. A lot of the output is spreadsheet-shaped: deterministic formulas, capacity constraints, data validation lists. If a route planner recipe computes cost-per-mile in a way that quietly ignores deadhead, that is your catch.

The other half is co-design. You help shape interview workflows the model uses to elicit requirements from a user, and you validate the context files it reasons over — dispatch logs, driver rosters, shipment manifests — for realism and internal consistency. Final quality sign-off on flagship recipes sits with you.

What the screen looks for

Specific operational history, not category familiarity. Fleet size, stop counts, service territory, what dispatch system you worked in, what you were accountable for.
Whether you can convert a dispatch rule into something deterministic. "Prioritise the nearest available tech" is not a spec; a tiebreak order, a skill-match condition, and a no-match fallback is.
Metric precision. Cost-per-mile versus cost-per-stop, what belongs in each denominator, when SLA adherence is measured against promise versus appointment window.
Evaluation judgment: can you distinguish output that is wrong from output that is merely unlike how you would have done it.

Prior AI evaluation experience is preferred, not required.

Logistics

Fully remote, contractor engagement, up to 10 weeks. Stated commitment is 40 hours per week with 8 hours overlapping PST; the listing notes part-time is acceptable but expects weekend on-call availability. You need your own desktop or laptop and reliable high-speed internet. Pay band is undisclosed on this listing — ask for the rate and the on-call terms in the Job Interest Form stage. Process is: Job Interest Form, then a delivery interview.

Operations / Supply Chain SME

What the work actually involves

What the screen looks for

Logistics

Interview prep — 10 questions