CAPABILITIES
Five practices. One operating logic.
Each capability below is a system I have architected, built, and run in production. They compose into a single operating model — but each can be engaged independently when the situation demands it.
01 — AGENTIC AI ARCHITECTURE
Multi-agent systems with deterministic guardrails.
I design agentic systems that survive production. That means hierarchical agent graphs with clear authority, deterministic guardrails around probabilistic components, audit trails on every action, and human-in-the-loop checkpoints at the points where stakes are highest. The frameworks evolve every quarter — the architectural principles do not.
02 — DATA & SEMANTIC PLATFORMS
Lakehouses with a semantic contract on top.
Modern data platforms fail at the same point: between the warehouse and the consumer. I design medallion lakehouses with a governed semantic layer above them — so analytics, ML, and AI agents all consume the same definitions of revenue, churn, and utilization, and no one is debating whether last quarter's number was correct.
03 — STRATEGIC DATA OPERATIONS
Treating data as a product, not a service ticket.
A data team that operates as a service desk will always be reactive and always be underwater. I install the operating model that turns data into a product line: pods with ownership, SLAs and SLOs on the things that matter, contracts on producer interfaces, deprecation policies, and a roadmap that the rest of the business can read.
04 — TEAM ORCHESTRATION
Designing teams that scale beyond 40 experts.
Past 25 people, a flat data team breaks. Past 40, a generic engineering org structure breaks too. Data and AI teams need a specific operating model — pods with mixed disciplines, an architectural authority outside the pods, and a delivery rhythm that does not collapse under a multi-region calendar. I have designed and run that model across three regions.
05 — EXECUTIVE ENABLEMENT
Turning leadership questions into operating decisions.
Data and AI investments fail at the executive surface, not the engineering surface. The dashboards are too many, the metrics conflict, the AI initiatives sound impressive but cannot be governed. I install the executive layer: a small set of decision-grade views, a clear AI governance model, and a quarterly cadence that lets a board or an executive committee actually steer the function.
ENGAGEMENT
How a working relationship begins.
Three modes, one bar for quality.
| Mode | Shape | Typical duration | Best for |
|---|---|---|---|
| Architect-in-residence | Embedded one or two days a week as the senior architectural authority. | 6–18 months | Companies rebuilding their data and AI foundation. |
| Executive advisory | Working with the CEO, CTO, or CDO on a recurring rhythm — strategy, hiring, governance. | 3–12 months | Leaders who need a senior peer to think with. |
| Diagnostic & rebuild plan | A focused 6–8 week assessment producing an architecture, operating model, and rebuild roadmap. | 6–8 weeks | Boards and PE operators evaluating an existing function. |
I take on a small number of engagements per year. Capacity is announced on the Now page. For specific availability and rates, the only path is a direct conversation.