Engagements.
Three ways to work with me: embedded in your team, with my ninja team, or in the room. Plus the case studies behind each one.
Three ways to work with me.
Senior fractional roles inside your team.
- Fractional Chief AI Officer
- Fractional AI Product Lead
- Forward Deployed AI Architect
Scoped delivery. Senior judgment, junior velocity.
- AI Opportunity Map
- Production Build
- AI Pod
Talks, workshops and training.
- Talks
- Hands-on workshops
- Team training
Me, embedded
Senior fractional roles inside your team.
You get me, part-time, at the table where the decisions get made. Pick the seat you are missing: strategy, product, or architecture.
Fractional Chief AI Officer
I own the AI roadmap and the systems that deliver it: what gets built, in what order, on what architecture, and whether it holds up in production. I am accountable for shipped outcomes, with the judgment to say "not yet" when that is the right answer.
- Weekly roadmap and delivery cadence with the exec team
- AI roadmap with phase gates: maximize what you already pay for, connect your systems, then build proprietary capability
- Governance framework a security team and a board can sign
- Vendor evaluation with a defined out-of-scope list, so we don't drift into vendor pitches
- Production scorecard: what shipped, what it costs, what it saves
Fractional AI Product Lead
For companies building an AI product without an AI product leader. I own the requirements, set the evaluation bar, run discovery and delivery in parallel, and work the security and change process that gets it launched. Technical enough to keep engineering honest.
- Product strategy and roadmap from MVP onward
- PRDs, user stories and acceptance thresholds for every AI capability
- Evaluation program: golden datasets, injected-error tests, blind SME review
- Backlog and sprint priorities across engineering, platform and vendor teams
- Launch: security and SQA reviews, UAT sign-off, release notes, leadership dashboards
Forward Deployed AI Architect
I sit with your team, design the architecture, write the code where it matters, and stay until the system runs in production. Senior enough to make the calls, hands-on enough to prove them.
- Architecture document with system diagram and cost model in week one
- AWS serverless reference implementation, deployed as infrastructure-as-code
- Evaluation harness that gates deployment before anything reaches production
- SSO, security review, observability and cost controls
- Pairing and code review with your engineers, then runbooks and handoff
Me + my ninja team
Scoped delivery. Senior judgment, junior velocity.
I lead and own the outcome. A small ninja team of AI master's students does the execution underneath me. You get an architectural adult in the room plus the throughput to actually ship.
AI Opportunity Map
A fixed-scope, board-ready diagnostic. Six questions a CEO or CAO actually has to answer, mapped against your operations, your data risk, and your team's ability to absorb new capability. Every finding ends in a dollar figure, an hour saved, or an explicit "not yet, here's why." Fully credited toward the follow-on engagement if we move forward inside 30 days.
- Current-state snapshot across departments or service lines
- Opportunity matrix scored on impact, feasibility, risk, time-to-value
- Top 3 plays, one-page brief per play with ownership plan
- Board-ready risk and governance brief (separate one-pager)
- 90-day roadmap, sequenced lowest-risk first, with a live readout
Production Build
A minimum lovable product, scoped to one customer or one workflow, built right the first time. AWS Lambda, Step Functions, EventBridge, DynamoDB, and Bedrock with structured outputs and prompt caching. Every LLM call wrapped in the deterministic envelope from the AI Production Framework, with replay and cost circuit breakers from day one.
- Architecture document with system diagram and cost model by end of week one
- Working pipeline shipped in vertical slices, deployed with CDK
- Evaluation harness gating deployment
- Build log, runbook, provider setup guide, replacement-path notes
- Handoff so your team can operate it without me in the room
AI Pod
When your engineering team is committed elsewhere, I stand up a self-contained pod outside it: me plus the ninja team, reporting straight to your CEO or product leader. We build the AI product, or find out whether you are ready for one, without pulling your engineers off their work.
- Pod charter: the goal, the scope, and how success is measured
- A working product shipped in slices, demoed to leadership every week
- Decision log and risk register leadership can see
- AI-readiness assessment (data, privacy, policy) where it matters
- A graduation plan: hand the product to your team, or keep the pod running it
I teach by building.
Talks, hands-on workshops and team trainings where people leave having built something, not just heard about it.

A hands-on build session for a full room of founders, operators and builders: real ideas from the room, taken from problem to working prototype with AI, live.
Talks
30 to 60 minutes on production AI for conferences, meetups and company all-hands: what it takes to get AI out of the demo and into the real world.
Hands-on workshops
Half or full day, in the 813 format. Everyone builds. The room brings real problems and leaves with something working and a method they can reuse.
Team training
Private sessions for the engineering, product and ops teams who build and run your AI: evaluation, agents and RAG, and the path from prototype to production.
Want your leadership team fluent with AI themselves? That is what I built Cobench for.
Recent engagements.
Each one tagged with the engagement type it proves. Names withheld by default. Reference calls available on request.
AI content review for a K-12 education company
Every lesson was reviewed by hand for grammar, style, bias and WCAG 2.1 AA accessibility, across 900+ courses a year. An earlier internal AI attempt had stalled with no evaluation protocol and no definition of done. The new tool had to read the authoring platform's JSON, plug into the company's AI course-delivery platform, and keep humans on final approval.
- •Scoped accessibility honestly first: mapped all 230 best practices in the reference WCAG library against what content JSON can actually prove, split into deterministic checks, AI-assisted checks and out-of-scope items
- •Set acceptance thresholds per check type and built the evaluation program: a 200+ example golden dataset, injected-error test suites and blind SME comparisons against Grammarly
- •Integrated with the course-delivery platform: API contract, scan callbacks, and automated Jira tickets for failed scans
- •Took it through enterprise security and SQA reviews, Entra ID SSO, and load testing at 200 concurrent users
- •Built a live reporting dashboard for QA leadership: issues by category and severity, dismissal trends, author drill-downs, cost by subject
On track to cut QA time per new course from 4 days to 2 hours and manual QA effort by 50%, with 90% detection at 5% false positives in evaluation and a cost of $0.37 per lesson.
AI SDR for a marketing analytics agency
The agency wanted outbound sales to scale without hiring more SDRs, and it wanted to offer AI to the same clients it already served with traditional analytics. Its engineering team was committed to client work, so the build had to happen in a self-contained pod.
- •Built the AI SDR on a durable agent runtime: ordering, idempotency, retries, replay, human approval and cost limits handled below the agent code
- •Lead, persona, mailbox and signal model with deliverability controls
- •Ran discovery sessions with the agency's clients to define what AI should actually do for them
- •Designed and stood up a new AI line of service alongside the traditional analytics offering
The AI SDR now runs sales conversations with 3,000+ prospects, and the agency has a new AI line of service to sell to its existing clients.
AI readiness pod for a medical group
A technology-forward medical group wanted to know whether it was ready for AI before committing to it. Referral coordination, with 30 people moving patients from primary care to specialists, was chosen as the test. The work had to stay inside HIPAA and the group's EHR and FHIR integrations, without pulling the main engineering team off its work.
- •Led a self-contained, five-person pod outside the main engineering team, owning both product and technical decisions
- •Built an AI-powered browser extension for referral coordinators, designed to cut time spent on referrals by 60%
- •Worked through data access, privacy and policy constraints directly with leadership
- •Delivered an AI-readiness assessment alongside the tool
A working referral tool plus an honest answer: data access, privacy and policy gaps had to be solved before more AI launches. Leadership put those fixes on its roadmap.
Signal-intelligence MLP for a sales-tech founder
Founder had a hypothesis: outreach triggered by life-event signals (new entity filings, real estate transactions, cross-state relocations) would outperform generic intent-based prospecting for tax and accounting services. Needed validation with real data inside a single client instance before investing in multi-tenant infrastructure.
- •AWS-serverless architecture: Lambda, Step Functions, EventBridge, DynamoDB, API Gateway, fully deployed via CDK
- •Signal pipelines for new entity filings (Cobalt Intelligence API + Florida Sunbiz parser + custom Virginia SCC scraper), real estate deeds (ATTOM Data with 133 Virginia jurisdictions), and cross-state relocation detection via buyer-previous-address cross-reference
- •Enrichment Step Function with Clay, Apollo, Hunter.io, plus email verification, so no unverified addresses reach outreach
- •Rule-based scoring with deduplication, suppression list, and configurable outreach-hold scheduling
- •Deliverables: architecture doc, build log, runbook, provider setup guide, daily output to HighLevel
Hypothesis testing pipeline running end-to-end on real Virginia + Florida filings and deeds, ready to validate response rates before any multi-tenant investment.
AI Readiness & Opportunity Map for a multi-site nonprofit
Chief Administration Officer holding the quality, consistency, and risk line across 10-plus centers and four service lines. The question wasn't whether AI was promising. It was whether AI could give 10 centers the consistency of one without adding headcount and without putting the families they serve at risk.
- •8 to 12 stakeholder working sessions across HR, IT, Quality, and three site directors across service lines
- •Site visits at two to three centers across early childhood and housing
- •Findings synthesized against four criteria: impact, feasibility, risk, time-to-value
- •Explicit "will not recommend yet" list: no client-facing AI, no AI in mental health delivery, no vendor recommendations
- •Six-question board-ready Opportunity Map plus a separate one-page Risk & Governance Brief for board, funders, and accreditation bodies
- •90-day sequenced roadmap, lowest-risk-highest-leverage first
Board-ready document delivered with a 90-minute live readout. Top three plays identified with named owners. The Chief Administration Officer left the engagement with a defensible answer to her board on what AI the org is and is not doing.
Chief AI Officer engagement inside a promotional products company
SAP Business One was the gravity center of the business. Nearly every operational pain point across finance, sales, client experience, and operations traced back to one of three SAP interaction patterns: getting data out, getting data in, or querying it. CFO had a 15-item automation priority list. The team needed a clear-eyed mapping of where AI belonged versus where deterministic automation was the right answer.
- •Five departmental audit interviews across Sales/CX, Finance, and Operations Leadership (including the CEO)
- •CFO's 15-item automation spreadsheet cross-referenced against transcripts to find universal pain points
- •Pain points mapped explicitly to AI Production Framework pillars: P1 (Design Doctrine), P9 (Economic Architecture), P10 (Operational Sovereignty)
- •Honest delineation between LLM tasks and tasks that should be scraping + templating + rules (the LLM Elimination Checklist applied)
- •Quantified ROI per pain point using real volume estimates (80+ daily SAP extractions in Finance alone)
Implementation roadmap with named pain points (PP-01 through PP-N), framework-mapped ownership, and a clear sequence that respects the team's bandwidth. The CFO's own "matching, accessing SAP and downloading, formatting" theme validated and prioritized.
Fractional Chief AI Officer at a 25-year creative agency
Mature agency with 25 years of domain expertise, enterprise clients, and a $1M internal platform already built. Not looking for a 90-day transformation pitch. Looking for someone who could get on the field, score quickly, and build toward something durable. The destination they were quietly working toward: a proprietary capability that encoded two decades of institutional knowledge as a competitive moat.
- •Phase 1: G Suite maximization. Gems for opportunity briefs, proposals, SOWs, creative briefs, meeting recaps, project handoffs, pre-mortems, milestone reviews. Notebook LM knowledge bases for delivery process and per-client context.
- •Drive governance: standardized folder structure mapped to the 5-phase delivery process, naming conventions, ownership rules so the library stays organized as it grows
- •Three live training sessions, run personally, not delegated. Recorded for future onboarding.
- •Phase 2 plan: HubSpot, Teamwork, and Slack integration, scoped separately
- •Phase 3 plan: proprietary MCP server encoding the creative process, client context, and methodology as a moat
Phase 1 designed with Phase 3's destination in mind, so nothing built in months 1-2 has to be undone when the agency builds its proprietary layer. Senior judgment plus junior execution velocity in the same engagement.
Permit-expediting AI for a single-operator construction-permitting business
Solo founder had already done the hardest part. 27 jurisdiction rules documents. One converted into a 13-page AI-ready reference with if/else logic. A Filipino VA with two years of construction permitting experience. A live Zoho CRM. A working cold email campaign. He was past the "figure it out" stage. He needed a tightly-scoped 6-week build that proved one jurisdiction worked end-to-end. A competitor had pitched LangChain + GPT-4 + generic Playwright at $1 to $2 per workflow.
- •AWS Lambda plus event-driven messaging instead of LangChain. Same orchestration benefits, no demoware fragility at scale.
- •AWS Bedrock for inference with model evaluation up front (not pre-selected). Fraction of GPT-4's token cost, validated via benchmark on real permits.
- •RAG + scripted chain-of-thought: the AI walks the founder's existing if/else jurisdiction tree step-by-step and shows its work, so the VA sees exactly where any wrong call happened
- •Doer/evaluator multi-agent pattern: one agent fills the permit, a second agent checks the first against the same rules doc before the VA sees anything
- •Tauri desktop app for the non-technical VA (free, included)
- •Enterprise serverless Playwright platform provisioned for Phase 3 portal submission (also included, separately valued at $30K+)
One jurisdiction working end-to-end. A playbook to onboard jurisdiction #2 without me. A reference benchmark on the engagement: a prior transcription framework I built priced at $0.0003 per transcript versus competitors at $0.30. 1000x cheaper, same quality. That's the cost discipline this build inherits.
What I've shipped.
Production systems I designed and built, from architecture to deploy, across six industries.
Easy Lemon AI
A law firm's platform for case-document management, intake and automated document workflows on AWS. I owned the product and the technical direction.
AI shipment-visibility platform
A carrier's own branded, AI-first product: ocean, rail and truck joined into one origin-to-depot journey. Multi-tenant React and AWS serverless; Bedrock writes the morning briefing and humans approve every action. 400+ commits.
Lead BOLDLY and the Anisa engine
A mobile AI leadership assessment on a live AWS conversation engine for a coaching firm, with per-coach tenancy, credit billing and Claude connectors. The employer pays and the executive owns the data.
Jobsite vision AI
Crews photograph a house and vision AI tells you its build phase (sheetrock, trim, finished and more) in any weather, then files it to the right community and lot. Built for gloved hands: two or three taps to capture and upload, offline capture that syncs when signal returns, and an admin app with per-lot timelines and crew analytics. A PWA on AWS serverless with Lambda, S3 and Bedrock.
Sports research platform
Tests strategy claims against real data before anyone acts on them: data pipelines, projections and simulation on AWS Fargate, a late-news agent, Claude (MCP) connectors and model evaluation scoreboards. 1,500+ commits.
Relationship AI SDR
An AI SDR that runs sales conversations for a marketing analytics agency, with human approval on outreach and cost limits on every run. 3,000+ prospect conversations and counting.
Durable Agent Runtime
The runtime the AI SDR sits on. Ordering, idempotency, retries, replay, human approval and cost limits handled below the agent code, so agents stay simple and production stays safe. 250+ commits.
Engineering runtime
A public marketplace of Claude Code plugins for engineering work done as planned, proven steps: audit a codebase, plan a rebuild a cheaper model can build, and run it as a reviewed, red-teamed loop.
AI Production Framework
The playbook behind every build on this page: 10 pillars and 140+ principles for taking AI from prototype to reliable, governed, cost-bounded production.