How Revolut Cut AI Coding Costs 90% With Self-Hosted Open Models - David Abitbol, Revolut

Why watch

David Abitbol from Revolut reveals how self-hosting open-source GLM models reduced their AI coding costs by 90% while achieving 60-70% faster delivery for 2,000 engineers. The talk covers GPU infrastructure requirements, latency targets that maintain developer satisfaction, tuning techniques that achieved 88% performance gains on the same hardware, and how an EU-regulated bank maintains compliance with self-hosted AI systems. Ideal for engineering leaders, platform engineers, and DevOps professionals considering self-hosted AI solutions.