Agents in Production: The Full Stack
Agents are getting better. The models are getting better. And yet teams are still shipping the same pile of workarounds they were shipping eighteen months ago. A vector store here, a graph database there, some glue in the middle, and a lot of... hoping for the best?
The problem has shifted from "can we build agents" to "how do we build agents that actually hold up." That's the conversation we're having in San Francisco on June 30.
We're bringing together engineers working at different layers of the stack - from the frameworks structuring agent behaviour, to the deployment and frontend layer, to the data substrate underneath it all.
Come with your production problems. We'll have food, drinks, and people worth talking to.
Agenda
17:30 - Doors + drinks + food
18:20 - Welcome
18:30 - Talks plus time for Q&A
19:45 - Drinks + networking
20:30 - Close
Talks
Tobie Morgan Hitchcock (CEO & Co-Founder, SurrealDB)
The Road to Agent Consciousness
We talk about agents that think, reason and remember. Strip away the demo and most of them have a context window and a prayer - they re-read a transcript each turn and hope the right thing is in it. Tobie takes the data-layer view of what it would take for an agent to hold something closer to a mind. He'll walk through Spectron, SurrealDB's memory and knowledge layer: memory that persists across sessions, reconciles what an agent is told against what it already knows, tracks what was true and when, and can always say where a belief came from. Cognition is a memory problem before it's a model problem - and he'll show it running live.
Priya Balgi (Senior Business Intelligence Engineer, AWS)
Why Your Agent's Data Layer Can't Be an Afterthought: Building Production AI Infrastructure for 20K+ Stakeholders
AI agents often get the spotlight, but the real challenge begins when you try to operate them reliably in production. Drawing from her experience building AI-powered systems used by more than 20,000 stakeholders at AWS, Priya explores the data infrastructure patterns required to move beyond prototypes and demos - covering data pipelines, evaluation frameworks, governance mechanisms, and the operational controls that help reduce hallucinations, improve reliability, and build trust in AI-generated outputs.
Bhrigu Srivastava (Software Engineer, Vercel)
The Connection Layer: How Agents Talk to Slack, GitHub, Notion, and Everything Else
Most production agents authenticate the same way: a long-lived, all-powerful API token in an env var. It's a security incident waiting to happen, and it doesn't scale past a couple of integrations. This talk covers building a secure connection layer between your agents to any 3rd party services without any long lived tokens. How agents can connect to Slack, GitHub, Notion, and any 3p service with Vercel Connect, using scoped, short-lived, delegated tokens.
About the speakers
Tobie Morgan Hitchcock (CEO & Co-Founder, SurrealDB)
Tobie Morgan Hitchcock is CEO & Co-Founder of SurrealDB. He is an experienced tech entrepreneur, developer, and software engineer, with 17 years' experience in the software and cloud-computing industries. He has a focus on distributed databases and highly-available architectures.
Priya Balgi (Senior Business Intelligence Engineer, AWS)
Priya Balgi is a Senior Business Intelligence Engineer at AWS who spends her time solving a problem many AI teams discover too late: reliable AI requires reliable data. Drawing on more than a decade of experience building large-scale data platforms and analytics systems, she focuses on the evaluation, governance, and infrastructure needed to make AI work in production. Her passion lies in turning promising prototypes into systems people can actually trust.
Bhrigu Srivastava (Software Engineer, Vercel)
Bhrigu Srivastava is a software engineer on Vercel's Marketplace & Connect team, where he builds the secure connection layer that lets apps and AI agents connect to third-party services. Before Vercel, he worked on large-scale distributed systems at Amazon, across the Prime Video and global logistics domains.
