Training & Agent Efficiency: a Snowflake NeurIPS Mixer

​Snowflake's AI Research team is building AI into the core of our platform. Our work spans training and inference systems, agent quality and efficiency, and contributes to open-source projects like DeepSpeed and vLLM.

​Join us during NeurIPS 2026 for an evening of technical talks, open Q&A and networking with fellow researchers, founders, and investors.

​Who should attend

• ​ML researchers and engineers interested in training systems and AI efficiency

• ​Founders and researchers at Neolabs and AI-native startups working on post-training, RL, inference or agents

• ​Investors focused on AI infrastructure and research-led companies

​Agenda

• ​5:00 PM | Arrivals and Networking

  • ​Drinks, bites, and conversation.

• ​6:00 PM | Cortex Training: Rethinking the Training Loop

  • ​Samyam Rajbhandari, Principal AI Architect, Snowflake (DeepSpeed co-founder)

  • ​How Cortex Training lets you keep control of the training loop while managed infrastructure handles the GPU-heavy work, and how multiplexing gets more out of each GPU.

• ​6:30 PM | CoCo: Quality and Efficiency in AI Agents

  • ​Anupam Datta, Principal Research Scientist, Snowflake (TruLens co-creator)

  • ​The research behind Snowflake CoCo: how the team measures agent quality and cuts token spend.

• ​7:00 PM | Networking

  • ​Keep the conversation going.

• ​8:00 PM | Close

​We're keeping the group small to make room for meaningful conversations. Secure your spot today!. Request your spot now!