Machine Learning in Production - The Hidden Effort

Explore the hidden engineering challenges of deploying machine learning models in production through this 13-minute conference talk from LeadDev Berlin 2025. Learn from 10 years of real-world experience at Chattermill about the substantial gap between AI demos and production-ready AI products. Discover the hard-won lessons on orchestrating dozens of self-hosted ML services, managing the unpredictability of LLMs at scale, and bridging the velocity differences between data science and engineering teams. Understand how seemingly innocent decisions like retries can result in hundreds of dollars in unexpected costs, and gain practical methods for driving down both processing time and infrastructure expenses. Master techniques for iterating faster and delivering more reliably in heterogeneous environments where data scientists and software engineers must collaborate effectively. Whether you're a tech lead, architect, or manager, acquire the knowledge needed to successfully ship real-world ML applications rather than just experiments, with specific focus on scaling, orchestration, and cost optimization strategies that separate successful AI projects from failed ones.