Engagements

AI assurance services for production AI.

Independent, evidence-grade assurance for US and Canadian teams shipping AI into revenue workflows. Fixed scope. USD. Senior IC only.

Currently accepting two engagements and one Officer seat. The Diagnostic is $7,500 and is credited in full toward any larger engagement within 60 days.

Book a Readiness Diagnostic

See the scorecard method

Specialist services

Go deeper on the buyer's problem.

Assurance

AI assurance services

Independent evaluation, evidence, and a board-ready readout for production AI.

Audit

Independent AI audit

Control testing, trace review, and NIST / ISO / EU AI Act readiness evidence.

Evaluation

AI evaluation services

k ≥ 3 evaluation for LLMs, RAG, and agents, with a harness your team owns.

Core

Start with evidence. Scale to assurance. Keep it current.

One ladder, one methodology. Every step produces the evidence the next step uses.

01 · Front door

Readiness Diagnostic

A scored read of one system, in one week.

Three axes scored, the top five findings, and a 30-day plan. Credited in full toward the full assurance or a fix sprint.

$7,500

Credited 100% within 60 days

1 week

Three of five axes scored
Top five findings with severity
One-page executive summary
45-minute readout with Q&A
30-day prioritized action plan

Best for A team with one system and one urgent question.

Credited 100% toward a Full Production Assurance or Fix Sprint within 60 days.

Book the diagnostic

02 · Flagship

Flagship

Full Production Assurance

A board-ready verdict on whether your AI is production-grade.

All five axes, 25 checks, repeated-run reliability, cost-per-success, and a control matrix. Includes evaluation-harness design and CI/CD installation, plus P0 remediation PRs. A 10–20 page report, a board deck, and a P0/P1/P2 roadmap.

$25,000

2 weeks

Five axes, 25 checks, k ≥ 3 runs
10–20 page report + board deck
Reliability and cost-per-success analysis
Evaluation harness design + CI/CD installation
P0 remediation PRs
Control matrix (NIST, ISO 42001, SOC 2, EU AI Act)
P0/P1/P2 remediation roadmap
60-minute readout + 30-day async

Best for A team scaling, raising, shipping to enterprise, or preparing for an audit.

Scope the assurance

03 · Recurring

Assurance Officer

A senior assurance lead in your corner, 20 hours a week.

Monthly scorecard, architecture review, eval ownership, and incident support. The evidence stays current between engagements.

$20,000/mo

90-day minimum · $24,000 after first renewal

20 h/week · 90-day minimum

Monthly scorecard update
Architecture and eval review
Incident support and postmortems
Quarterly board summary
Async Slack response within 24 hours
Vendor and model evaluation
Weekly evaluation and regression review

Best for A team that needs ongoing evidence and senior judgment.

Apply for the seat

Specialist

Regulated systems, remediation, and investor diligence.

The same methodology, scoped for a specific decision with a different audience.

Specialist · Regulated

Regulated Assurance

Technical evidence a notified body or enterprise reviewer will accept.

The full assurance plus risk classification, AI Bill of Materials, model and dataset cards, and a control matrix mapped to NIST, ISO 42001, SOC 2, and the EU AI Act. Legal review through a partner.

Best for EU-exposed or regulated AI providers preparing for conformity assessment.

$50,000

3–4 weeks

Discuss regulated assurance

Specialist · Remediation

Assurance-Driven Fix Sprint

Fix the P0s the assurance found.

One system, 4–6 weeks. Two tracks: retrieval/cost or agents/reliability. The work lands in your repo with an eval suite in CI and a re-score that proves the improvement.

Best for A client with an assurance report or an incident.

$40,000

$35,000 founding-client rate for the first two clients

4–6 weeks

Scope the fix

Specialist · Investors

AI Technical Due Diligence

An independent technical read before the deal closes.

A 1–2 week assessment for VCs, PE, acquirers, and insurers: model differentiation, data provenance, reliability evidence, inference cost, vendor dependency, team capability, and red flags. A 10–15 page memo and an IC readout.

Best for VC/PE, acquirers, and insurers.

$15,000–$30,000

per deal · partner rate for repeat flow

1–2 weeks

Discuss a deal

Recurring

Evidence expires. Re-score it.

AI systems drift, vendors change models, and last quarter's evidence stops being true. A yearly re-score keeps the record current.

Recurring · Annual

Annual Re-Assurance

Keep the evidence current.

A one-week re-score that checks drift, vendor changes, and regression. Updated evidence pack and a renewal recommendation.

Best for Every client after an Assurance or an Officer year.

$8,000–$12,000/yr

Scope-dependent

1 week/year

Schedule the re-score

The method

Five axes. Twenty-five checks. One index.

Every engagement is scored on the same standard. The index is a decision threshold, not a grade: it tells you what to fix before you scale, raise, or face an auditor.

30% weight

Reliability & Eval Coverage

Does it work consistently, and can you prove it across runs?

20% weight

Cost-per-Success & Routing

What does a successful outcome cost, and when should traffic route cheaper?

20% weight

Guardrails & Adversarial Safety

What happens when it is wrong, attacked, or misused?

15% weight

Observability & Drift

Can you reconstruct what happened and detect drift before a customer does?

15% weight

Regulatory Control Readiness

Can you prove it to an auditor, a board, or a regulator?

Critical 0–39 Fragile 40–59 Production-capable 60–74 Assured 75–89 Exemplary 90–100

The full scorecard and all 25 checks are public. Read the model, the bands, and what each check asks for before the fit call. Read the scorecard →

How an engagement runs

Three predictable phases. No surprise invoices. No scope creep.

Week 1
Scope
Start with the self-serve 5-Axis Scorecard. The 20-minute SOW alignment call uses your score to define one system, the axes in scope, access, success criteria, and the delivery date. No code changes until it is signed.
Weeks 1–2
Assess
The harness runs the golden set k ≥ 3 times. Stakeholder interviews, trace review, and document review fill the gaps. Every check gets a score and an evidence link.
Week 2
Readout
A 10–20 page report, a board deck, and a P0/P1/P2 roadmap. The 60-minute readout is a decision meeting, not a sales pitch. Fix work is optional and separately scoped.

Anti-promises

What I refuse.

Six lines that protect the work. If any of these disqualifies you, this page just saved us both an hour.

01
I refuse hourly billing.
Fixed scope or a monthly retainer. Always.
02
I refuse unguided brainstorming calls.
Every conversation starts with evidence: run the self-serve 5-Axis Readiness Scorecard or request a 90-second async repository teardown before we scope a $7,500 Diagnostic.
03
I refuse more than two concurrent engagements.
Quality compounds with attention, not headcount.
04
I refuse NDAs before a paid Diagnostic.
After the fee clears, I will sign yours.
05
I refuse junior hand-offs.
I take your engagement, I deliver every line.
06
I refuse agency subcontracting or marketplace work.
Direct contracts only. No markup, no middle layer.

If any of this ruled you out, we saved each other an hour. If it didn't, book the diagnostic.

Common questions

If yours isn't here, ask on the fit call.

What exactly is production-AI assurance?

An independent, evidence-based assessment of an AI system that is already built or about to ship. It covers reliability, cost-efficiency, safety, observability, and governance, and produces a scored index plus a prioritized remediation plan. It is not a certificate or a generic code review; it includes architecture and design evaluation, repeated-run testing, harness implementation, and scoped remediation.

How long does an engagement take?

The Readiness Diagnostic is one week. Full Production Assurance is two weeks. Regulated Assurance is three to four weeks. A Fix Sprint is four to six weeks. The Assurance Officer is a monthly seat with a 90-day minimum.

What does it cost?

Diagnostic: $7,500, credited toward a larger engagement. Full Assurance: $25,000. Regulated Assurance: $50,000. Fix Sprint: $40,000, with a $35,000 founding-client rate for the first two clients. Officer: $20,000/month for 20 hours/week. Annual Re-Assurance: $8,000–$12,000/year. Due Diligence: $15,000–$30,000 per deal.

Do you work with US and Canadian teams?

Yes. The practice is remote-first from Mumbai with dedicated US and Canada overlap (ET/PT). Cal.com shows your local time. Invoices are in USD; USD and CAD settlement are available through Wise, and Deel can handle the contract where required.

What is the payment structure?

Fixed-scope engagements are 50% on signature and 50% on delivery, Net-15. The Officer bills monthly in advance. Every invoice is in USD; USD and CAD settlement are available through Wise, and wire/Stripe are accepted where supported. Late invoices carry 1.5% per month.

Who owns the IP?

You own the report and any code merged into your repository. I retain the methodology, the scorecard, the harness, and generic templates. The SOW states this explicitly before work starts.

Will you sign an NDA?

Yes, after the paid Diagnostic is confirmed. I do not sign NDAs before a paid call; the fee is what keeps the advice independent and the calendar sane.

Why not just use an eval tool or a public benchmark?

Tools produce data; they do not produce an independent judgment. Public benchmarks measure a model in a lab, not your system in production. The scorecard measures your system, on your tasks, with repeated runs and a cost baseline, and gives you a document you can show a board or an auditor.

Start here

One week. One scored read.

Book the Readiness Diagnostic: $7,500, credited toward any larger engagement within 60 days. If the evidence says you do not need us, you still leave with the findings.