An event dedicated to exploring all things AI Engineering!
Event partner: Brex
🕒 5:30 pm – 6:00 pmWelcome and Opening Remarks
🎤 Marijan Cipcic, Principal Events Manager at Daytona
🕒 6:00 pm – 6:10 pmTalk "Forkable Sandboxes for Long-Running Agents"
🎤 Muhammad Annas Hashmi, DevRel at Daytona
Outline:
Agents are starting to do real software work: installing dependencies, editing files, running tests, starting servers, and debugging over long sessions. But most AI systems still treat each run like a disposable prompt-response loop. That breaks down once the useful state lives not just in tokens, but in the actual machine.This talk is about the runtime layer agents need to become reliable builders. I’ll show how forkable Daytona sandboxes let you pause at important moments in an agent run, branch the environment, and sample multiple continuations from the same machine state. For builders, this turns long-running agent workflows into something you can inspect, compare, recover, and eventually optimize.
🕒 6:10 pm – 6:20 pmTalk "TBA"
🎤 Connor Guerrero, Sr. Developer Relations Manager at Crusoe
Outline:
TBA
🕒 6:20 pm – 6:30 pmTalk "Building production-ready agent integrations"
🎤 Bastien Beurier, co-founder and CEO at Nango
Outline:
Connecting an agent to an API is easy. Making it work across real users, accounts, and permissions is harder. I’ll share a practical playbook for agent connectivity across three main dimensions: coverage, reliability, and governance. We’ve used this playbook to power integrations for teams like Replit, Mastra, and ServiceNow.
🕒 6:30 pm – 6:40 pmTalk "Good evals are boring"
🎤 Lotte Verheyden, Developer Relations Engineer at Langfuse
Outline:
Long-running agents fail in ways that are hard to spot, and with traces alone this becomes tedious to monitor. This talk covers how to turn traces into evals that serve you in improving your agent's execution quality.
🕒 6:40 pm – 6:50 pmTalk "How to Build Truly Episodic Memory for Your Agent"
🎤 Liz Zhang, Developer Relations Lead at FalkorDB
Outline:
Most people build their agent's memory on a vector index, but similarity and memory are not the same thing, so top-k retrieval can wrongly drop the record that explains what happened, mix two different experiences, or return a fact that is no longer true. In this talk, I show you how to model an episode as context, action, outcome and evidence, with time and people as real relationships, and why "retrieve more" and metadata filters only partly fix the problem. You leave with a practical way to build episodic memory for your agent that truly remembers you and your experiences.
🕒 6:50 pm - 8:00 pm
Networking
With pizzas and beverages
About event
This is dynamic gathering for AI enthusiasts, innovators, and professionals to collaborate, share ideas, and explore the latest advancements in artificial intelligence. Whether you're building AI products, researching cutting-edge algorithms, or simply passionate about the field, join us to connect, learn, and drive the future of AI forward.