Engagement model
A week-by-week rubric your board can hold us to.
Not a 40-page roadmap. Not a six-month pilot with three checkpoint meetings. Six weeks from scoping call to production handoff, with a cancel-and-refund gate at week 2.
Week 0
01
Scoping call with a senior engineer
60 minutes with the engineer who would own the build, not an account manager. You leave with a written one-pager: the specific production outcome we would own, the data sources it touches, the eval rubric, and a fixed fee.
- →Outcome: one specific shipped feature, not a 40-page roadmap
- →Pricing: fixed weekly rate, published up front, no ramp-up fee
- →Exit: you can cancel after week 2 if the prototype misses its rubric
Week 1
02
Engineer onboarded to your repo
The named engineer is in your GitHub, Slack, standup, and feature-flag system by day 2. We ship the first real PR in 7 days or we pause billing until we do. No 3-week diligence phase.
- →First PR: within 7 days of start
- →Access: read/write on the repo we would change, no more
- →Standup: daily async update in your Slack, on your timezone
Week 2
03
Prototype running in your staging
Behind a feature flag, hitting your real data, with your eval harness attached. Measured against the rubric from week 0. If we miss the rubric, the engagement ends with prorated refund and the code stays with you.
- →Running on your stack: Bedrock / Vertex / Azure / OpenAI / Anthropic, your call
- →Eval harness: ragas plus a case-specific rubric, in your repo
- →Decision gate: continue to production or refund + exit
Weeks 3-5
04
Production hardening
Eval scores stabilize, p95 latency hits target, audit trail lands, FinOps dashboard shows per-token spend. Your team pair-programs with ours so the handoff is not a cliff.
- →Observability: PostHog / LangSmith / Datadog / OTEL, whichever you run
- →Cost: per-token budgets with alerts, documented fallback model
- →Compliance: EU AI Act readiness checks, audit log, red-team passes
Week 6
05
Handoff and departure
Runbook, architecture doc, failure playbook, eval harness, and CI/CD are all in your repo. We run a 90-minute handoff session with your on-call rotation, then we leave. The engineer stays available for paid 2-hour consults at a published rate for 12 months after.
- →Leave-behind: runbook, eval, architecture doc, failure playbook
- →Ownership: you own the IP, the weights layer, and the repo
- →12-month option: paid 2-hour consults with the same engineer, capped rate
Written into the MSA
Four commitments, in the contract, before week 1.
First PR in 7 days
Or billing pauses until we ship.
Week-2 prototype
On your rubric, or we refund and exit.
Named engineer
The person you meet in week 0 is the person writing the code.
Your IP
In writing, in the MSA, before week 1 starts.
Scope one outcome. Ship it in six weeks.
The scoping call is free. You leave with a written plan whether you hire us or not.
Book the scoping call