​What if you could build your next big feature without anyone looking at it? 

​Claude Code, Cursor, and Codex are very good at editing files in your terminal session.

​When it comes to running autonomous agents, it becomes harder.

​You need durable state, scoped credentials, isolated runtimes that can run your actual stack, audit logs, cost control, and more. You also need to know whether the agent is any good for your product, not just whether it scores well on a public benchmark.

​We are going to talk about different ways to build and evaluate autonomous coding agents.

​Engineers and tech leads building with coding agents. Free, limited seats.

​Agenda

• ​17:00–17:30 Welcome

• ​17:30–18:00 Talk 1

• ​18:00–18:30 Talk 2

• ​18:30-19:00 Break

• ​19:00–19:30 Talk 3

• ​19:30–20:00 Cheers

​Talks

​Talk 1: How to build AI Agents that connect to sensitive data

​By: Alon Gubkin, CEO, Alien.dev

​Most AI demos work because the data is easy to reach. Real enterprise data is not.

​It lives in private databases, internal APIs, customer VPCs and other systems security teams will never expose to the public internet.

​In this session, we'll build an AI worker that runs inside a customer-owned environment while the AI agent stays in your cloud. The agent does the reasoning; the worker performs approved actions near the private data, like reading files, querying systems, and writing results back, without opening inbound ports or copying raw datasets upstream.

​We'll show how to deploy the worker and how this architecture works across AWS, GCP, Azure, and on-prem environments, without opening inbound ports or asking customers to export raw datasets upstream.

​

​Talk 2: Agents that fix CI, review PRs, and spin up previews

​By: Assaf Ben Josef, Senior Software Engineer, Islo

​Most PR review agents read diffs, but they don't run your app.

​Assaf built autonomous agents that take CI failures and fix PRs, review PRs and spin up PR preview environments. They spin up real MicroVMs with the full system, verify the fix actually works, review the change in a live environment, and publish a PR preview.

​In this session he'll walk through what they built, what broke along the way, and the infrastructure that makes it possible.

​Talk 3: Evaluating Agents for Your Product

​By: Nir Naim, AI Lead | Autonomy AI

​Generic benchmarks don't tell you much about whether your agent handles your stack, your failure modes, or your customers' edge cases.

​Nir will cover Agent-as-a-Judge patterns and how to build an eval suite for your product - golden cases, scoring rubrics, synthetic and production replay - and where to run it (CI, pre-release, post-incident).

​Who Should Attend

​Engineers and technical leads building with or on top of AI coding agents - whether you're just getting started or already running agents in production.