Building reliable agents: a practitioner's playbook
About this session
Agents fail in ways that traditional services do not: partially, plausibly, and expensively. This talk is the playbook we wish we had — bounded tool surfaces, idempotent actions, replayable traces, checkpointing, and the escalation paths that keep a human in the loop without keeping them in the way.
Speakers
Sofia Marchetti
Founding Engineer
Vantage Labs
Sofia was the first engineer at Vantage Labs and now leads their agent runtime. She is unreasonably interested in what happens when a tool call fails halfway through.
Ava Nakamura
Staff ML Engineer
Lumen AI
Ava leads the evaluation platform at Lumen AI, where she is responsible for making sure model changes ship without quietly regressing customer-facing quality. She spent six years in search relevance before moving to LLM systems.