Bay State Data Book a 30-minute call

Principal-led AI advisory · Boston

Before your AI costs you real money, get a senior second opinion.

Bay State Data is a principal-led practice. Every engagement is led, reviewed, and signed by Andrew Greenhut — an internal-audit-trained AI evaluator — at fixed, transparent prices, with no incentive to sell you a build.

Book a 30-minute call

A scoping call is not a commitment. You'll get a straight read on whether an engagement makes sense.

You spent money on AI — is it B.S.?

Nobody independent and senior is telling you whether it works — what it will cost when it scales, what happens when it breaks, or whether the vendor's claims hold up. Your vendor grades their own homework. Your team is heads-down and too close to it. And hiring a $250K AI lead just to get a second opinion is the wrong tool.

Bay State Data is the anti-vendor. No platform to sell you, no partnership incentives, and a hard rule: we never sell a build before an independent look at whether you need one. When we find a problem, we tell you what it is and what it costs you — and the fix is scoped from evidence, not from what's convenient to sell.

Diagnosis before prescription. Fixed, transparent pricing. Senior review on everything.

Three engagements. Fixed prices. No open meter.

AI Opportunity Assessment

$6,500 fixed · 1 week

For companies that haven't built AI yet. Everyone's selling you AI — where would it actually pay? A structured diagnosis of your business: we walk your workflows, find the candidates, screen them for feasibility and data readiness, and hand you a prioritized roadmap.

  • A prioritized use-case roadmap with success measures defined before anything is built
  • A "don't build this" list, with reasons — the part no vendor will give you
  • Portable: take it to any builder, including not us
  • Feasibility grounded in your actual data and systems, not vendor demos

Fixed price. Independent. Diagnosis before anyone sells you a build.

Start with a scoping call

AI Systems Audit

$9,500 fixed · 1 week

An independent senior second opinion on your production or near-production AI system — vendor-built or in-house. A structured 12-module review methodology, run with staged gates and internal-audit discipline.

  • Evidence-bound, severity-ranked findings — no claim without proof
  • Board-readable report: clear for leadership, specific enough for your technical team
  • Prioritized remediation roadmap you can act on the week it lands
  • Findings walkthrough with your team

Fixed price. No open meter. No surprise scope.

Start with a scoping call

Build Sprints

Fixed quote per sprint

When the roadmap calls for building or fixing, we deliver it — as scoped milestones, not open-ended hours. My team executes; I scope every sprint and sign off on everything before it reaches you.

  • One named deliverable per sprint, with acceptance criteria agreed up front
  • Fixed quote and delivery date before work begins — we manage how it gets done
  • Out-of-scope requests become the next sprint, never a change-order fight
  • Typically scoped from an audit roadmap, so estimates come from evidence

Fixed price. Senior-reviewed. Scoped from the roadmap, never sold before it.

Start with a scoping call

Need something narrower? A Focused Review (a chosen subset of the audit's dimensions) and a standalone Technical Review are available as scoped-down engagements, priced at scoping.

How the audit is built

The AI Systems Audit is a structured 12-dimension review with staged gates and evidence-bound findings — informed by NIST AI RMF and ISO/IEC 42001 principles, delivered with internal-audit discipline.

Every claim in the report is bound to evidence we examined. Where evidence was withheld or unavailable, the report says could not verify — we never fill a gap with a plausible-sounding story. Each finding carries a severity, the business impact in leadership terms, and a recommended action.

Two honest limits: the methodology synthesizes recognized frameworks — it is not a certification, and we don't claim one. And we flag risk; we don't render legal opinions — that's your counsel's call.

How it works

  1. 1 · A 30-minute scoping call

    You describe what you have and what you're worried about. We tell you directly whether an engagement makes sense and what it would involve — before either of us commits to anything.

  2. 2 · A fixed-price engagement

    Every engagement starts with a named deliverable, a fixed price, and defined acceptance criteria. Scope is locked in the engagement letter before work begins. No hourly ambiguity, no open-ended scope.

  3. 3 · A straight answer

    You get findings you can take to your board and a roadmap you can act on immediately — with every claim bound to evidence we examined.

Recent work

Selected engagements, anonymized per NDA.

  • Public-safety nonprofit — senior-led ML audit and production deployment hardening: shadow-scoring rollout, AWS infrastructure review, XGBoost/FastAPI stack.
  • Private investment buyer — recurring AI technical audits: independent review of AI-related claims in diligence contexts.
  • Data pipeline engagement — NLP and OCR data-pipeline engineering: end-to-end pipeline design and validation.
  • Ecommerce company — semantic search overhaul: OpenSearch, k-NN and hybrid retrieval, catalog-scale deployment.

Who you're working with

Andrew Greenhut, President of Bay State Data

I'm Andrew Greenhut. Bay State Data is a principal-led practice: I lead every engagement — scoping, diagnosis, synthesis, and the final report — and my team and I conduct the hands-on technical review. You have one accountable senior point of contact, and nothing reaches you without senior review.

My background is the evaluator's, not the vendor's: trained as a Certified Internal Auditor (CIA) with 3+ years of internal-audit practice — planning, fieldwork, evidence, findings — and formerly PMP-certified. Two engineering degrees from MIT (BS '06, MS '10) and fifteen years in production ML across data science, auditing, and IT consulting: ML and text analytics in production defense environments at Raytheon, live fraud-detection and sales-scoring models at GoTo, and five years as a Product Director at DataRobot working with teams deploying AutoML in the field.

The auditor's instinct is the product: I judge whether the people who built your AI knew what they were doing — and whether they can prove it.

Get a straight answer about your AI.

Start with a 30-minute scoping call. You'll know whether an engagement makes sense — and what it would actually involve — before either of us commits to anything.

Book a 30-minute call