Agentic civic AI for a U.S. state’s residents, federal public safety web at NHTSA, and Amtrak booking redesign. Institutional constraint as design brief.
- Governed agents for state services
- NHTSA
- Amtrak
$10.5M pre-seed · governed agents · a civic pilot under active stage gates
Government & Civic · Product direction + design · 2015–2026
The institution is the constraint. The citizen is the brief.
Government technology fails at the last mile, not because the capability doesn’t exist, but because no one resolved which workflow could absorb the change before procurement locked in. That pattern runs across a decade of public-sector work.
NHTSA (2015–2017) was federal public safety communications: recalls, ratings and defect information, all made readable without being made wrong. Amtrak (2017–2019) was redesigning national booking as a system, not a form flow, at 30M+ traveler scale. Neither involved AI. Both required the same discipline: absorb institutional complexity, don’t expose it to the citizen. On governed agents for state services (2024–2026), the work extended to agentic AI on a state government’s citizen-facing mobile channel: methodology, wedge definition, and governed agents under structured stage gates.