ML Pub Club #27: Building Enterprise AI for High-Risk Customer Communication: Control Layers, Agents, and Escalation in Production

​At the next ML Pub Club, we’re going deep into what enterprise AI actually looks like when it runs in production - not in demos, not in benchmarks, but in regulated, high-risk environments where mistakes are expensive.

​In telecom and financial services, an LLM is never “the system.” It’s one component inside a carefully designed control architecture built for correctness, auditability, and trust.

​In this session, we will unpack real production patterns: explicit control layers, coordinated agent groups, persistent state management, and deterministic escalation paths. We’ll also explore failure modes that don’t show up in model benchmarks - but surface very quickly in real-world deployments.

​If you’re building AI systems that need to work under pressure, this is a practical engineering session focused on closing the gap between model performance and operational reliability.

​The session will be led by Juraj Malenica, Head of Engineering at Mindsmiths, where he leads the design of large-scale AI systems for enterprise customer engagement. With more than a decade of experience in distributed systems and backend engineering, Juraj focuses on system architecture and production reliability. He co-leads the development of the Mindsmiths AI Platform, serving millions of users across Europe and the US in regulated, high-stakes environments.

​🗓️ When: Tuesday, March 3rd 2026 | 18:00
📍 Where: CroAI HQ (Zavrtnica 17, 4th floor)
🔗 Registrations via Luma

​Join us at ML Pub Club for a candid, engineering-first look at enterprise AI in production.

​See you there!