Strategy often stops at the handoff. I close that gap by designing the operating model, directing the systems and infrastructure, and staying accountable to the production result.
One accountable operator, from direction through production.
The evidence is deliberately separate: changing how a large human organization operates, and creating an AI-native system that turns direction into shipped software.
As COO of a 1,200+ person technology group, I led an operating-model redesign that reduced operating expense by approximately CHF 30 million and increased throughput by 65%.
For my own companies, I designed a system that lets one operator deliver like a team, with grounded context, autonomous dispatch, and quality controls built into the path to production.
Autonomous ticket-to-pull-request dispatch turns direction into reviewable change. A three-voter adversarial gate challenges every AI output before it passes.
One half proves organizational change. The other proves a new production model. Together, they show one operator building the machinery of an engineering organization, then using it to ship across radically different domains.
The work begins with a consequential operating problem and ends with a dependable system, clear controls, and an organization that can keep using it. Rigor is concentrated where failure is expensive: deterministic runtime tests, signed command paths, RLS and security boundaries, recovery systems, and approval gates.
Choose where AI can create material value, sequence the work, and keep accountability with a human operator.
Design the platform, context layer, security boundaries, verification gates, and production infrastructure.
Deliver a working system on real company data instead of stopping at a strategy deck or prototype.
Codify the skills, governance, approval paths, and operating rhythm that let the system expand safely.
The agent fleet is the method beneath four proof domains: industrial real-time systems, simulation, legal services, and consumer mobile. The domain changes. The discipline does not.
Industrial real-time automation platform
Deterministic runtime, a physics-based digital twin, an 18-module operator console, and on-prem ML. Directed and shipped before any engineering hire.
Agentic delivery infrastructure
A system that lets one person deliver like a team, with quality controls built in so the output is trustworthy, not just fast. Four MCP servers provide semantic search, code and symbol indexing, infrastructure-state collection, and browser verification.
A second company estate, directed by the same operator
Mission planning and simulation studio, common operating picture with live terrain, ISR tooling, and real-time simulation backends across 8 deployed properties.
Legal-services platform, closed beta, onboarding clients now
Zero-trust access administration with expiring, revocable, sign-in-limited access codes. Access codes are stored only as SHA-256 digests. The database is ringfenced with no public port.
Consumer couples app, coming soon
The current build is live on the App Store and Google Play. It comprises 90k lines of Dart, with self-hosted Supabase, row-level security throughout, compressed delivery, and automated CI.
The pilot is the smallest useful proof. Larger engagements build the system, controls, and operating capability around it.
$1,500 / 72 hours
One working automation on your real data.
Start the pilot$4,500+ / 1 to 2 weeks
A production agent workflow with verification gates.
Scope the build$5,000+ / 1 to 2 weeks
A dashboard or console on your data, with auth.
Scope the tool$10,000 / 2 weeks
Map where AI can ship, then leave a working pilot and governance playbook.
Scope the auditNeed a broader operating mandate?
Scope a larger engagement
My path runs from US Navy Intelligence, through ventures and a COO turnaround, to designing the machinery of an AI-native engineering organization.
I direct the system and remain accountable for its output. The machines multiply capacity. Experience, taste, decisions, and judgment determine what should be built and where rigor belongs.