VP of Public Sector @ aiXplain · Co-founder of ThinkHive (acquired by Rootly) · Founder of OmiraTech
I've spent my career on one problem from different angles: making autonomous systems trustworthy enough to carry real weight - agents in production, benefits people can claim, government decisions a person can inspect and contest.
- VP of Public Sector at aiXplain - building the public sector practice for the operating system of autonomous work (45K+ models, deployed across 63 countries, sovereign and air-gapped environments included). Bringing agentic AI to federal, state, and local government held to the bar that setting demands: traceable, auditable, provable.
- Previously: Member of Technical Staff, National Design Studio (White House) - shipped projects that changed how Americans interact with their government under the America by Design initiative. Learned there that when government software works, a right on paper becomes a right people can actually hold.
- Co-founder of ThinkHive (acquired by Rootly, July 2026) - the AI agent reliability platform: traces every step an agent takes, correlates metrics/traces/evals to catch hallucination and drift, clusters failures into patterns, and validates fixes with shadow testing before they ship. Now part of Rootly, the incident management platform trusted by NVIDIA, Replit, and Canva.
- Founder of Omira Technologies - US engineering studio (Chicago, est. 2018) engineering AI into products, machines, and operations. Eight years of shipping for clients who needed it to actually work.
- Agentic AI in high-stakes environments - evaluation, observability, drift detection, shadow testing
- Reliability engineering applied to LLMs - SLOs for agents, regression gates, groundedness scoring
- Public-interest technology - working public technology is how a state keeps its promises
- Production systems over demos - enterprise and government will only deploy what you can prove
| Project | What it is | Stack |
|---|---|---|
| hallucination_detector | Deterministic LLM hallucination guardrails - pluggable detectors, severity escalation, CLI + Python API | Python |
| MLFlowX | Lightweight DAG engine orchestrating production ML/AI workflows, with tracking dashboard and plugin system | Python |
| model-forge-cli | CLI for managing and running AI models locally | TypeScript |
| omira-code-cli | Developer CLI with AI-powered code generation and project management | TypeScript |
| SalmonNet | Deep learning for salmon run forecasting | Python |
Languages: TypeScript · Python · Rust
Focus: Agentic AI · Agent reliability & evaluation · MLOps · GovTech · Next.js · FastAPI · Kubernetes
Something new is cooking. Same problem, much bigger surface area - you'll hear it from me soon.
Working on AI in the public sector - federal, state, local, or serving them as a vendor? My inbox is open: LinkedIn