Pricing

Three ways to engage.
Pick what aligns with your goal.

No retainers, no inflated SOWs. We propose the model that puts our incentives next to yours — and we're transparent about what each one costs.

One-time project fee

Best for: a specific outcome with a defined scope. RAG system in 6 weeks. SOC 2 audit prep. Cost-cut sprint.

$60k – $400k+

depending on scope · fixed price

  • Written scope, fixed deadline
  • Architecture & ROI doc up front
  • 50% on kickoff, 50% on acceptance
  • Eval-gated acceptance criteria
  • Knowledge transfer & runbooks included
  • 30-day warranty
Discuss a project →

% of savings

Best for: high-confidence cost optimization on inference, GPU spend, or cloud waste. We only get paid if you do.

25–35% of verified savings

12-month measurement · third-party audited

  • Free 2-week opportunity audit
  • Written savings forecast and method
  • Baseline locked at engagement start
  • Monthly savings reports, your billing data
  • You keep 65–75% of savings forever
  • Cap on payout if you want certainty
Run a free audit →
Which one fits?

A quick decision matrix.

If you need… Project Hourly % Savings
A defined deliverable on a fixed deadline★★★★★
Embedded engineering for ongoing work★★★
Strategic advisory or architecture review★★★★★
Cloud / inference cost optimization★★★★★★★
Compliance prep (SOC 2, HIPAA, EU AI Act)★★★★★
No internal AI / ML platform team yet★★★★★
Pricing FAQ

Questions we get often.

Do you do retainers?

No. Hourly is our flexible model — same billing rhythm as a retainer, but you only pay for hours used. We send a monthly summary and adjust together.

How is "savings" measured for the percentage model?

We lock a baseline at engagement start (last 90 days of cloud / inference bills, usage-normalized). Savings = baseline run-rate − new run-rate, measured monthly, reconciled quarterly. We can bring in a third party to audit if you'd like.

Do you sign NDAs and customer DPAs?

Yes — both, before any technical conversation. We also carry $5M E&O insurance and can sign vendor security questionnaires (SIG, CAIQ).

Where do you deploy code? In your accounts or ours?

Yours. Always. We get scoped, time-bounded IAM access. We never run production workloads in MLOPS-owned accounts.

What if we want to bring it in-house later?

Great — that's the goal. We document, train your engineers, and run a structured handoff. Many clients move from project → hourly → "call us when something breaks" within 18 months. We consider that a win.

Not sure which model fits?

Tell us what you're trying to do — we'll recommend the engagement model that aligns our incentives with your goal. No commitment.

Book a 30-min call