Building CI/CD for non-deterministic AI agents at scale

Details
AI agents break traditional CI/CD assumptions: same input doesn't guarantee same output. How do you test, deploy, and monitor systems that behave differently each time? This talk covers practical patterns for building reliable deployment pipelines for non-deterministic AI agent systems.
Attendees will learn:
- Why traditional CI/CD patterns fail with AI agents and what new failure modes to expect (inconsistency, ‘it worked on my machine’, problems due to model changes, etc.)
- What are evals (evaluations) and how you can use them to drive improvements to your AI Agent
- How to deploy evals in CI/CD pipelines and measure the impact of every change to agent quality
After a 45-minute talk there’ll be a 15-minute Q&A, for which we encourage you to submit questions in advance.
A webinar recording and related materials will be shared with all attendees after the event.
___________________
Speaker:
Natan Yellin - CEO @ Robusta Dev
Natan Yellin is the founder of Robusta Dev. Before founding Robusta he was a senior engineer and long-time open source contributor to projects like GNOME Linux. When not working on software, Natan can be found in the garden with his son, growing fruits and vegetables for their family.

Building CI/CD for non-deterministic AI agents at scale