Events
Engineering Reliability: A Practical Playbook for AI Evals
Meetup photo 1

Building production-ready AI requires more than just a great prompt. It requires an evaluation strategy. Christos Zigkolis, who leads Applied AI Engineering for enterprise and agentic systems at Pfizer, joined us to demystify AI Evals and the shift from traditional deterministic testing to probabilistic systems. We explored the taxonomy of Evals, defined clear team roles, and discussed how evaluation-driven development makes non-deterministic AI systems more reliable, measurable, and ready for enterprise use.


Through a live agentic RAG use case, Christos showed how to move beyond “vibe-based” testing and systematically evaluate performance at scale. The first hour covered the fundamentals and concepts behind AI Evals, followed by hands-on material that gave attendees the opportunity to experiment with the showcased techniques and tools themselves.

Meetup repository on GitHubPractical-AI-Thessaloniki-Meetup/meetup-5
(opens in a new tab)