

InferenceCon
The AI revolution isn't just about training anymore – it's about inference at scale. With industry projections showing inference costs set to dwarf training expenses by 10x by 2027, understanding how to optimize inference is becoming critical for every ML team.
Join BCV for an exclusive gathering that cuts through the noise: InferenceCon. We've assembled an exceptional group of practitioners who are solving real inference challenges today:
Charles Frye from Modal Labs will share battle-tested techniques for inference optimization at scale, drawing from his experience helping companies train and deploy models efficiently.
Charlie Snell from UC Berkeley will dive into ways to scale test-time compute optimally, and how it can be more effective than scaling model parameters.
Daria Soboleva from Cerebras Systems will talk about how DeepSeek's models leverage MoE (Mixture of Experts) architecture, and the challenges that come with scaling them at inference.
Through focused lightning talks, you'll hear about practical strategies you can implement immediately. As inference becomes the dominant cost center in ML operations, this event will facilitate a deep technical dialogue with other engineering leaders and practitioners tackling similar challenges.
Space is limited; come prepared with your toughest inference optimization questions for our speakers. Light refreshments will be served.