

Daytona AI Builders - SF, October '25 @Datadog
โAn event dedicated to exploring all things AI Engineering!
โAgenda
โโโโโ๐ 5:30 pm โ 5:40 pm
Welcome and Opening Remarks
โโโโโ๐ค Marijan Cipcic, Principal Events Manager & Asako Hayase, Pacer, Community Ambasador at Daytona
โ๐ค Max Andaker, Principal Product Manager at Datadog
โโโโโ๐ 5:40 pm โ 5:50 pm Lightning Talk "How we built Open SWE: An Open-Source Asynchronous Coding Agent"
โโ๐ค Brace Sproul, Head of Applied AI at LangChain
โโโOutline:
โOpen SWE is an open-source asynchronous coding agent that autonomously plans and executes software changes across entire repositories. Built with LangGraph, it writes and edits code running on Daytona sandboxes. Open SWE combines structured planning, human-in-the-loop collaboration, and automated execution that takes an idea, or GitHub issue, and turns it into a production ready pull request.
โ๐ 5:50 pm - 6:00 pm
Lightning Talk "From LLMs to SLMs - automated fine tuning, distillation and routing with Datawizz"
โโ๐ค Iddo Gino, Founder, CEO & Data Labeler at Datawizz
โโโOutline:
โLLMs are great for prototyping, but hard to scale with. Inference cost, accuracy, latency and reliability can hurt applications. In this talk we will show (live demo) how to use the OSS datawizz platform to identify use cases, distill SLMs and set up smart model routing. We will go beyond the basic examples - using vision data to show for to tune SVLMs and share some best practices from existing deployments!
โ๐ 6:00 pm - 6:10 pm
Keynote "Orchestrating Cloud Code: Parallel Sandboxes with Daytona"โ
โ๐ค Vedran Jukic, Co-Founder & CTO at Daytona
โโโOutline:
โDaytona is an advanced AI sandbox platform delivering lightning-fast, secure infrastructure for running and testing AI-generated code. This demonstration showcases how Daytona Sandboxes transforms Cloud Code into a powerful orchestration platform for distributed development workflows. Using a local Cloud Code instance as the orchestrator, we'll demonstrate how natural language prompts can spawn and manage multiple parallel cloud instances running in isolated sandbox environmentsโenabling scenarios like parallel test execution across different configurations or coordinated multi-step development tasks.
โ๐ 6:10 pm - 6:20 pm
Lightning Talk "What AI Agents Can Build When They Understand Video"
โโ๐ค Amy Xiao, Founder & CEO at Founder at Cloudglue
โโโOutline:
โWhat if your AI agent could watch a coding tutorial and instantly recreate the app? What if it could watch a stream of ads, and build an evolving trend dashboard - without humans?
โCloudglue turns videos into first-class data sources by giving AI agents native video understanding. For the first time, you can go from video โ interactive experience, or chat across hundreds of videos as naturally as working with text.
โIn this demo, we'll show an AI agent that uses Cloudglue to โwatchโ a coding tutorial, build a playable HTML5 game directly from the video, and deploy it as a working browser game all on its own. With Daytonaโs secure sandboxes, the app runs safely and reproducibly in isolation, with dependencies and assets installed and managed.
โItโs a glimpse of a world where video is no longer just consumed, but now queried, built upon, and turned into living, interactive experiences.
โโ๐ 6:20 pm - 6:30 pm
Lightning Talk "Receiving webhooks in your AI agent"
โโ๐ค Tom Hacohen, Founder & CEO at Svix
โโโOutline:
โAgents are already immensely capable, but they can only do good work if they have the right context and can react to events as they happen. Webhooks are the common way to enable that. Though while webhooks are easy to use in most scenarios, they are not always easy to consume in the context of agents. This demo will cover an alternative way to consuming webhooks with agents.
โโ๐ 6:30 pm - 6:40 pm
Lightning Talk "Using Apify and Daytona to safely test code from the Internet"
โโ๐ค Frantiลกek Nesveda, Senior Engineering Manager at Apify
โโโOutline:
โApify is a web scraping and automation platform offering thousands of ready-made scrapers for various usecases, providing tools and context for your AI applications. In this demo, we will use Apify to scrape untrusted code from the Internet and safely test it in Daytona's isolated sandbox environments. We will achieve this without writing a single line of code, utilizing Apify's and Daytona's MCP servers to control the two platforms. This enables rapid exploration, prototyping and testing of untrusted code and new algorithms without worrying about security or infrastructure.
โโโโโโโโโโ๐ 6:35 pm - 8:30 pm
โโNetworking
โWith pizzas and beverages
โโ________________________
โโAbout event
โDaytona AI Builders Series brings together the people building the next layer of AI infrastructure. Each month we focus on practical problems for scaling agents โ from secure execution and cost control to reliability and new protocols. The goal: equip builders with the tools and insights to ship agent-native systems into production.