Inferact: Building the Infrastructure That Runs Modern AI

Woosuk Kwon and Simon Mo, creators of vLLM and founders of Inferact, talk with a16z's Matt Bornstein about how modern AI models are served in production, why inference became one of the hardest AI infrastructure problems, and their vision for a universal open-source inference layer. Video version: https://www.youtube.com/watch?v=GsRnarLIC9g