https://www.youtube.com/watch?v=Ok6QdZHCcI8The presentation explains how Revolut transformed its engineering productivit...

https://www.youtube.com/watch?v=Ok6QdZHCcI8The presentation explains how Revolut transformed its engineering productivity and AI costs by self-hosting open-source AI models, particularly GLM, on dedicated GPU infrastructure. By giving its 2,000 engineers effectively unlimited access to #AI Revolut reports a 60% reduction in coding time, a 70% improvement in engineering lead time and approximately 90% lower AI costs compared with commercial providers. A key lesson is that optimising AI infrastructure and the wider engineering workflow matters more than simply buying additional GPUs, with Revolut achieving an 88% improvement in model-serving performance without increasing its GPU capacity. The company uses a hybrid approach, reserving frontier models for specialist tasks, while emphasising security, sandboxing, human oversight and regulatory compliance. Overall, the presentation argues that self-hosted AI can become a cost-effective, scalable part of enterprise engineering infrastructure...