I architect regulated systems — healthcare, security, finance — and stay through compliance, deployment, and traction. The diagram, the controls, and the board story, from one set of hands.
CISSP · HIPAA audits cleared first pass · agent graphs & eval gates in production · graph-backed retrieval · causal inference
Fig. 01 filters the record below. Try it.
Every claim traces to a capability — the same way a control traces to evidence.
I harden systems until audits are boring: controls architecture, threat models, PHI-grade security designed in from the first diagram — so audit season is a formality and a breach isn't an extinction event.
Why me: CISSP — the credential that seats a CISO · HIPAA audits, first pass · billing rails secured
I take the workflows burning your team's hours and hand them to agent graphs — with human gates exactly where judgment matters. Headcount goes back to the work only people can do.
Why me: agentic intake in production · graph & loop engineering · evals that gate what ships
I point causal and active inference at your data estate — agents that forage for the information they lack and find where value hides, instead of waiting for a dashboard to confess. Discovery becomes a system, not a lucky analyst.
Why me: 2 yrs population-scale modeling · causal + active inference stacks · LLM systems in production
I build the system and carry it through procurement, partner review, and production. Deployment is the deliverable, not the epilogue — if it isn't adopted, it didn't happen.
Why me: clinical dashboards clinicians open · enterprise partner builds shipped · systems still running
Two years of production data science on GLP-1 therapeutics at population scale, reported into public-health channels. Called elevated progression risk for obesity-associated comorbidities ahead of the literature — the finding held. Being early only counts if you're also right.
Designed the technical controls and carried deployments through HIPAA audit — cleared on the first pass. Compliance treated as an architecture input from day one, not a remediation project at the end.
Built and secured healthcare billing systems — where PHI, money movement, and audit trails intersect and there is no partial credit for almost-correct.
Clinical coaching dashboards running on live care workflows — data products clinicians actually open, because adoption is the only metric that matters at the surface layer.
HIPAA-compliant agentic intake for a federal-benefits law practice, integrated with CMS Blue Button — an agent graph with consent rails and human gates at every determination.
Architected and shipped builds inside large corporate partner structures — where the technical design has to survive procurement, security review, and two org charts at once.
Technical due diligence for investors and acquirers, and standing advisory for executives — translating between the codebase, the controls, and the valuation model without losing anything in transit.
Agentic LLM systems as an engineering discipline — planner and critic graphs, convergence loops, eval gates deciding what ships. The graph work predates the agent wave: DAGs and graph-shaped pipelines were the job for years before orchestration frameworks made them fashionable. Everyone arrived at loop engineering recently. I was already here. In production, not in prototypes — forward-deployed inside venture builds on Sand Hill Road, where my job is getting healthcare ventures through the compliance and security gates that are the real moat.
An agentic sourcing and market-intelligence platform running inside a venture studio: automated discovery, enrichment, and scoring across 3,500+ companies and still growing, alongside the investor organizations and accelerator programs they emerge from. Discovery, enrichment, and evaluation as separate layers — with the investment-readiness rubric held as configuration, never as code, so the thesis can change without touching the engine.
Causal models to learn what actually moves the needle; active-inference agents that seek out the information they lack. The aim: automate the finding of value across data projects — not just the reporting of it.
Earlier theaters: e-commerce pricing, market risk, production data science across finance. The methods travel; the constraints just get sharper.
Every architecture decision is a capital decision. I've sat on the C-suite side of the table — valuation, financing, cost of delay — and design with that ledger open.
Policy used to be a document beside the system. In an agentic stack it's context — and context is runtime behavior. I compile the controls into the build: enforcement at retrieval, versioned prompts, audit trails that were never bolted on. Audits pass the first time because there's nothing to reconstruct.
Decks don't hold load. I stay through deployment and adoption, because a system that isn't carrying real traffic is still a hypothesis.
A regulated system has to work, be allowed, know things, and be true. That's usually four vendors and no one accountable — the value is in the seams no single specialist can cross.
An inventory of the AI agents already running in your environment: what they can reach, what they're authorized to do, what an auditor would find, and what to fix first. A findings memo your board can read and your engineers can act on. Most teams discover they can't answer question one.
Start hereForward-deployed inside your team — your repo, your standup — taking a regulated system from concept through compliance to production load.
Start hereTechnical due diligence for a deal, or standing counsel for a board or C-suite that needs the codebase translated into capital terms.
Start hereCISO engagement: controls architecture, audit preparation, HIPAA-native design — trust built into the system, not appended to it.
Start hereIntros, collaborations, engagements, or a hard problem you want a second architect's eye on — all welcome. The best messages are one paragraph: what's stuck, what it costs, what you'd do with the capacity.
I read every message personally — you'll hear back within one business day.