Course artifact
Design partner offer
A fill-in-the-blank pilot offer covering baseline, target, inputs, commitments, commercial terms, and end states.
Design partner offer
Workload
One model, request shape, environment, and owner.
Frozen baseline
- Quality:
- p95 / p99 TTFT:
- Inter-token latency:
- Cost per successful 1M tokens:
- Deployment lead time:
Target
One primary measurable outcome plus non-regression constraints.
Inputs from partner
Representative traffic, model artifacts, evaluation corpus, weekly owner access, and security review.
Our commitment
Four-to-eight-week staffed pilot, weekly report, incident response, final benchmark, and production recommendation.
Commercial
Pilot fee, included capacity, overage, data handling, IP, reference conditions, and conversion decision date.
End states
Expand to production, extend once for one named gap, or stop.