Agent systems,
built by someone
who knows why they break.

Strategy, architecture, and honest evaluation for teams building serious work in AI. Fifteen years designing autonomous systems, from DARPA-funded research to shipping products. Open for new engagements.

01 / 04

Agent Architecture & Strategy

System design for multi-step, tool-using, goal-directed agents. Where should the loop live? When does planning beat reaction? What's the smallest thing that could actually work? I help you answer these before you build the wrong thing.

System design Build vs. buy Evaluation harnesses
02 / 04

Research to Product

Translating cutting-edge AI research into systems that ship. I read the papers, know the tradeoffs, and have seen what collapses in production. For teams that need someone who can talk to both the researchers and the engineers.

Prototyping Technical scoping Roadmaps
03 / 04

Proposal & Grant Review

DARPA, SBIR, and government AI proposals. I've been on the submitting side ($5M+ awarded) and know what evaluators respond to. Pre-submission review, technical narrative, experimental design, risk framing.

DARPA / SBIR Technical narrative Red team review
04 / 04

Technical Due Diligence

Honest assessment of AI claims, for investors, acquirers, or boards. What's real, what's vaporware, what's "real but won't scale." Independent technical judgment from someone who has no product to sell you.

Investor DD M&A review Board memos
15yr
Building AI agents
since 2011
$5m+
Prior research
funding
30+
Peer-reviewed
publications
Past & present
DARPA-funded PI Naval Research Laboratory Lehigh University Indiana University Bloomington OpenSats-funded

Metacognition. World models. Long-duration autonomy. Knowledge representation. Fifteen years on the problems where the right architecture matters more than the right model.

I do my best work on problems that don't have a clear playbook yet. New agent architectures, research translating into products, evaluation for systems that are hard to evaluate. If you're in that territory, I'd like to hear about it.

Typical engagements run from a half-day strategy session to multi-month architecture work. I'm comfortable with NDAs, classified-adjacent contexts, and the administrative texture of government programs.

  • Defense & government AI programs
  • AI-native startups past the prototype stage
  • Enterprise teams standing up agent infrastructure
  • Investors & boards evaluating AI companies
  • Research labs commercializing their work

Let's talk about what you're building.

Available · Replies within 48 hours