Agent Harnesses and Systems Engineering
Details
PRESENTATION OVERVIEW
Agent harnesses are becoming the way teams turn nondeterministic AI into something worth trusting in real work. This talk gets into what it actually takes — the engineering patterns, the failure modes that only show up after many runs, and the design choices that matter most — based on agentic AI work in security and infrastructure. Trust in agents isn't a property of the model. It's something you engineer around it.
Speaker:
Bryan Starbuck
Principal AI Research Engineer with 25+ years of experience, focused on building agentic AI systems for critical infrastructure. The interesting problem in agentic AI right now isn't capability — it's reliability over long horizons. My work centers on the scaffolding around models: verification agents, adversarial testing, and regression frameworks that enable empirical trust-building in agents over time.
Event Details:
- Location: 1951@Skysong @ Scottsdale, AZ
- Time: 5:30 PM - 7:30 PM - talk will begin at 6:00pm (we get kicked out promptly at 7:30pm)
- Free Parking: Easily available (free)
- Food: Simple Pizza and Water