​An event dedicated to exploring all things AI Engineering!

​Event partner: Postman

​​Agenda

​​​​​​🕒 5:30 pm – 5:40 pm
Welcome and Opening Remarks

​🎤 Marijan Cipcic, Principal Events Manager at Daytona

​​​​​​​​🕒 5:40 pm – 5:55 pm
Talk "Teaching a Weak Code Model to Improve with Reinforcement Learning"​

​🎤 Daniel (Thi) Graviet, Machine Learning Engineer at Daytona

​​​​​​​​Outline:

​What happens when you give a small, weak code model a way to learn from its mistakes?

​In this live demo, we'll use Group Relative Policy Optimization (GRPO) to fine-tune a code model through trial and error. The model will generate solutions to coding tasks, run them in isolated Daytona sandboxes, and receive rewards based on whether its code passes the tests.

​We'll walk through the full loop—from generating completions and grading them in parallel, to feeding the rewards back into training and watching the model improve over just a few steps. Along the way, we'll explore why reinforcement learning for code is challenging, how sandboxed execution makes automated feedback possible, and what it takes to make these systems reliable enough for a live stage demo.

​You'll leave with a practical look at:

​- How GRPO trains models without traditional labeled answers
- Using Daytona sandboxes as safe, parallel code graders
- Designing simple, explainable rewards for coding tasks
- Observing a model's performance improve in real time

​​🕒 5:55 pm – 6:10 pm
Talk "Agents Are Your Teammates : Building and Deploying Agents with Astro"

​🎤 Pooja Mistry, Senior Developer Advocate at Postman

​​​​​​​​Outline:

​Work is changing shape. Instead of doing everything yourself, you can now delegate to AI agents you hire, train, and monitor, agents whose effects compound your own success. This is happening not just in engineering, but in marketing, sales, product, and support. Drawing on real examples from building and running agents on Astro AI, this talk explores what it takes to train and manage an agent teammate, and what we've learned about the economics of running agents at scale. In this session, you'll learn how to step by step build and deploy your own teammates, right where you work.

​​🕒 6:10 pm – 6:25 pm
Talk "SKILL Issue: Inside the Attack Surface of Agentic Skills"

​🎤 Nipun Gupta, Founder & CEO at Optimus Labs

​​​​​​​​Outline:

​Every agentic product ships with a new kind of dependency: skills — natural-language instruction bundles that enter a model's context before any human reads them and execute helper code with full user privileges. If your product publishes skills, auto-installs community ones, or persists agent memory, congratulations: you are now a package registry.

This talk maps that surface end to end. First, anatomy: what actually executes when an agent "learns" a skill, and why the lethal trifecta (private data × untrusted content × external egress) makes the format uniquely fragile. Second, the just-published **OWASP Agentic Skills Top 10** - from malicious skills and supply-chain compromise through insecure metadata, weak isolation, update drift, and poor scanning — each risk anchored to a real 2026 incident. Third, the uncomfortable state of the **spec**: Anthropic maintained agentskills.io defines file shape but nothing downstream enforces it i.e. no signing, no permission model, no provenance, while OWASP's Universal Skill Format proposal shows what trust could look like if registries adopted it. Fourth, the receipts: how badly is the supply chain already hit? One marketplace flooded with 1,184 malicious skills in days; a brand-impersonation campaign clearing every scanner at 300K+ installs per skill; 12% of live skills resting on attacker-controllable external instruction sources; infostealers specifically hunting agent identity files.

You'll leave with the shared vocabulary (AST01–AST10), a builder's map of which risks you personally create per role i.e. publisher, embedder, harness operator, runner, and the defense stack sized to what detection can actually catch today.

​​🕒 6:25 pm – 6:40 pm
TTalk "Running Data Agents on Enterprise Systems"

​🎤 Brian Ramprasad, Co-founder of VantEdge Labs

​​​​​​​​Outline:

​Enterprises want AI agents that can analyze their data—but handing untrusted, model-generated actions on live systems like Salesforce or the data warehouse is a non-starter for security and compliance teams. Solving that safely is what unlocks data agents inside the enterprise. This talk covers how VantEdge builds analytical agents on top of Daytona: agents pull data from systems like CRM and financial data sources, then run their analysis right inside the sandbox using our lightweight SDK. No credentials are ever injected, so even compromised code has nothing to steal. We'll walk through a real workload end to end, from ingesting data to producing an agentic data product, and show how the entire process remains controlled and auditable.

​​​​​​​​​​​​​​​​🕒 6:40 pm - 8:30 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.