Shipping
Jun 19Testing Platformknowledge_v2 baseline + ADR 0015 (MEDIC) build gate opens. Fri AM ran the first…Jun 19Testing PlatformFirst GUARDIAN-auto-surfaced legacy bug (F1). Fri AM: while diagnosing the OPI Classic…Jun 19Testing Platform4 knowledge_v2 prompt lessons encoded + verified stuck on first regen. Diagnosed Wed's 7…Jun 17Testing PlatformConfidence scoring fix — knowledge_v2 drafts can now bucket HIGH (commit 7117861).…Jun 17Testing Platform4 follow-on fixes from real-data SME ingest session (commit b89fa5e). (1) nginx 1MB…Jun 17Testing PlatformTESTGEN preview/expand UX (commit 08a800d). Two follow-on bugs spotted after knowledge_v2…Jun 17Testing PlatformKNOWLEDGE → TESTGEN loop closed end-to-end (commit 0dc4b1c). New knowledge_v2 generator +…Jun 17Testing PlatformKNOWLEDGE ingestion accepts files (not just paste-in text) — commit fc74962 + 62fccf1.…Jun 17Testing PlatformGUARDIAN Inbox drill-in restored — commit 99f3e6d. Tuesday's flow-first reshape removed…Jun 17Testing PlatformADR 0019 drafted: server-side exception ingestion (ELMAH → CODEX automatic). Replaces…Jun 19Testing Platformknowledge_v2 baseline + ADR 0015 (MEDIC) build gate opens. Fri AM ran the first…Jun 19Testing PlatformFirst GUARDIAN-auto-surfaced legacy bug (F1). Fri AM: while diagnosing the OPI Classic…Jun 19Testing Platform4 knowledge_v2 prompt lessons encoded + verified stuck on first regen. Diagnosed Wed's 7…Jun 17Testing PlatformConfidence scoring fix — knowledge_v2 drafts can now bucket HIGH (commit 7117861).…Jun 17Testing Platform4 follow-on fixes from real-data SME ingest session (commit b89fa5e). (1) nginx 1MB…Jun 17Testing PlatformTESTGEN preview/expand UX (commit 08a800d). Two follow-on bugs spotted after knowledge_v2…Jun 17Testing PlatformKNOWLEDGE → TESTGEN loop closed end-to-end (commit 0dc4b1c). New knowledge_v2 generator +…Jun 17Testing PlatformKNOWLEDGE ingestion accepts files (not just paste-in text) — commit fc74962 + 62fccf1.…Jun 17Testing PlatformGUARDIAN Inbox drill-in restored — commit 99f3e6d. Tuesday's flow-first reshape removed…Jun 17Testing PlatformADR 0019 drafted: server-side exception ingestion (ELMAH → CODEX automatic). Replaces…

State · County · Municipal · Mission-Driven

AI modernization for public-sector teams under real constraints

I help state, county, and city agencies turn accessibility mandates, legacy workflows, and institutional knowledge risks into secure, maintainable AI-enabled systems — with measurable outcomes, clear governance, and documented handoff. Currently embedded as the AI lead for Multnomah County, running 5 production initiatives that any procurement officer can verify.

5

Production AI initiatives

56.5K

County files monitored nightly

92.3%

Independent audit agreement

704

SME sessions captured

Where I Help

What I do for public-sector teams

Accessibility & WCAG Automation

PDF-to-HTML pipelines that clear backlogs before federal ADA Title II deadlines. axe-core validation, deterministic forms, human review queue.

Site Intelligence & CMS Auditing

Nightly Drupal-aware audit of files, pages, accessibility, SEO, links, PII, performance. Validated against independent external audits.

Legacy Modernization Support

Behavioral test capture from retiring SMEs. Acceptance contracts that make rebuilds safe. Knowledge corpora with full provenance.

AI QA & Testing Platforms

Multi-agent systems that index codebases, generate regression tests from captured knowledge, and guard against drift on every commit.

Secure Internal Knowledge Systems

RAG over agency data with role-based access. PHI guardrails. Agency-controlled cloud — zero data leaves the environment.

Executive Advisory & Training

Fractional AI leadership, governance frameworks, change management, executive workshops, and staff enablement programs.

When I'm The Right Call

Why agencies bring me in

Compliance deadline approaching

ADA Title II (April 2027), Section 508, state accessibility mandates, or records modernization deadlines you can't manually meet.

Staff capacity gap

Critical knowledge held by 1-3 SMEs nearing retirement, or AI workload that won't scale without specialist hires you can't get budgeted.

Failed AI pilot

Proof-of-concept worked but the system never made it to production. Procurement, governance, or technical handoff stalled the rollout.

Legacy system risk

Custom-built application running for 15-30+ years that nobody fully understands, and the rebuild estimates from vendors all start at $5M.

Manual process overload

Tagging, classifying, remediating, or reviewing tens of thousands of records by hand — at a pace that will never catch up.

Need for secure internal automation

AI workflows that can't be sent to a third-party SaaS because of data classification, PII, or jurisdiction constraints.

How I Work

Three engagement options

Most engagements start with Discovery — a short fixed-fee assessment that scopes the embedded engagement (or honestly tells you it isn't the right time). No surprises, no sales-cycle bloat.

Tier 01

Discovery

2–3 weeks · $8K–15K fixed

Stakeholder interviews + workflow walk-throughs
Top-1-or-2 opportunity sketch with architecture
Plain-language deliverables for executive review
Recommended roadmap + sizing for the embedded engagement (or honest 'not the right time' if that's the answer)
Flagship

Tier 02

Embedded AI Program Lead

12–24 weeks min · $25K–40K/month retainer

Own the AI program end-to-end — architecture, code, adoption
Ship features weekly with a dated delivery record (the same kind you can see at the Multco hub)
Capture institutional knowledge before SMEs retire
Stakeholder hub, executive briefings, ROI tracking
Train your team to run it without me by the end

Tier 03

Architecture & Build Sprint

4–10 weeks · $30K–80K fixed

Specific system architected, built, and deployed
Multi-agent pipelines, RAG, deterministic forms parsers
Production deployment + observability baseline
Hand-off documentation + train-the-trainer session

Procurement Snapshot

How to contract with me

The information a procurement officer needs to start an internal conversation about an engagement, in one place.

Cossette Consulting LLC · Entity & Procurement Detail

Legal EntityCossette Consulting LLC
Entity TypeTexas Domestic LLC (filed Dec 29, 2022)
TX SOS File #0804858459
TX Taxpayer #32087716570
Federal EIN92-1502035
Texas StatusRight to Transact Business: ACTIVE · SOS Registration: ACTIVE
NAICS Codes541511 (Custom Programming) · 541512 (Systems Design) · 541618 (Management Consulting) · 541690 (Scientific & Technical Consulting) · 611430 (Training)
Engagement TypesDirect PO · Retainer · Fixed-price project · Sole-source justification · Subcontractor under cooperative purchasing primes
Cooperative Purchasing PathsAvailable via Sourcewell, OMNIA Partners, TIPS, and NASPO ValuePoint primes — happy to identify a vehicle that maps to your existing procurement procedures.
InsuranceGeneral Liability, Professional Liability (E&O), and Cyber Liability — current Certificates of Insurance available on request.
LocationSan Antonio, TX · Remote primary · On-site as needed

Proof Of Work

The Multnomah County AI Program

Five concurrent initiatives. One embedded architect. A live presentation hub with a dated delivery record that any procurement officer or peer agency can verify in 60 seconds.

Live Program Hub

Verify the work in real time at the Multnomah County AI Development Hub — each initiative has its own page with architecture, stats, and a dated activity feed.

multco-presentations.web.app
Dated delivery recordIndependent third-party validationZero data leaves county cloudDocumented handoff in every engagement

Ready to talk

Bring me in to lead your AI program.

Start with a 30-minute working session — not a sales pitch. We map what you're trying to solve, I tell you whether embedded AI program leadership is the right fit, and you get a written summary you can forward internally.