Work
I do two kinds of work. I advise institutions moving AI from experimentation to production, drawing on a bench of specialists — engineering, regulatory, security — assembled per engagement. And I build working systems, each one a specific, falsifiable hypothesis about where AI and financial infrastructure are going. The systems generate the operational knowledge the advisory work runs on.
Advisory Practice
AI Strategy for Financial Services Leaders
I work with private equity-backed enterprise software companies, banks, and payment institutions to move AI from experimentation to production. The practice is built on three decades of infrastructure experience: architecting Visa’s real-time payment systems and serving as CTO at a digital asset bank through one of the most volatile regulatory periods in modern finance.
Engagements are designed for senior leadership — executives deciding where to invest in AI, how to govern its deployment, and how to build advantage without breaking regulatory obligations.
Engagement types
Executive AI Strategy Briefings
Half-day sessions for C-suite and technology leadership. Competitive landscape, regulatory framework, use case prioritization, and 90-day action planning.
Governance Framework Development
AI risk committee charters, model risk management approach, vendor accountability frameworks. Designed for board presentation and regulatory examination.
Implementation Advisory
Ongoing strategic support during the first 90 days of AI deployment. Vendor evaluation, use case scoping, board communication strategy.
Team AI Enablement
Hands-on working sessions for engineering and product teams adopting AI tools and building with GenAI APIs.
Working Systems
Each system starts as a hypothesis: a statement about the near future that could turn out to be wrong. I build it to find out. What survives contact with real users and real constraints becomes advisory material; what doesn’t gets retired.
Finance Meets Agentic AI
What has to exist before agents can move money.
KnowYourAgent
Trust Infrastructure for AI AgentsHypothesis
As agents begin initiating payments, institutions will spend more proving agents can be trusted than making them smarter.
A verification and trust platform for autonomous agents in commerce and financial services: verified agent identity, real-time merchant verification at sub-100ms latency, and immutable reputation scoring. Treats agent identity as a financial primitive, the way KYC treats human identity.
StablecoinRoadmap
Payment Infrastructure for Programmable MoneyHypothesis
Banks will adopt programmable money through product validation, not policy debate. The institutions that can simulate stablecoin flows first will ship first.
A platform for building and launching stablecoin payment applications: working templates for wallets, gateways, and remittance platforms, with sandbox simulation for validating payment flows before committing to rails. Built on operational experience navigating stablecoin regulation at a digital asset bank.
Agent Infrastructure
How agents get evaluated, governed, directed, and pointed at systems that predate them.
Versalist
AI Evaluation PlatformHypothesis
Static benchmarks will not predict production performance. Challenge-driven evaluation — real systems under real constraints — will become how institutions assess AI capability.
A challenge platform where engineers build and evaluate working AI systems, from single tasks to multi-agent architectures. The evaluation methodology it generates feeds directly into my assessment frameworks for AI in high-stakes financial decisions.
AutoDevOps
AI Agent Governance for the Regulated SDLCHypothesis
Coding agents will reach regulated engineering organizations faster than the evidence trail their examiners require. What stalls the rollout is not the model — it is the inability to prove what the agent did.
Captures what Claude Code, Cursor, and Codex do inside the SDLC — every session, tool call, and decision — and turns it into evidence an examiner can verify: policy that allows, confirms, or blocks an action before it runs, an append-only audit log in storage the institution controls, and evidence packages that verify offline without exposing raw prompts or source.
ActsAsGeek
Agentic Workflow OrchestrationHypothesis
The scarce skill in agentic work is command, not construction. A single operator will direct workloads that once required a team.
An agentic command center for orchestrating, monitoring, and scaling AI workflows as a solo operator. A live study in what directing autonomous workers actually requires — the coordination patterns apply directly to financial operations.
MigrateForce
AI Migration EngineHypothesis
Enterprise AI will not be bottlenecked by models. It will be bottlenecked by the legacy systems models cannot reach.
A high-speed, AI-driven engine for enterprise application migration. Tests whether agentic systems can compress multi-year modernization programs into months — the precondition for AI adoption at institutions running decades-old cores.