Samyam Rajbhandari

Snowflake AI Research is focused on delivering high-quality, efficient AI for real enterprise workloads — bridging state-of-the-art research with practical product impact. In this session, we’ll share what’s new across our research agenda, highlighting key directions including model-agent co-design to improve task success and reliability, vertical models specialized for enterprise use cases, and advanced training and inference systems that enhance usability and performance while reducing cost and latency. We’ll also share principles and lessons learned, and how research transitions into production to help teams build trustworthy AI applications on governed data at scale.

Discover the latest innovations from Snowflake AI Research across foundation models, agentic technologies and inference systems. Learn how Arctic Text2SQL reasoning models set a new standard in accuracy with best-in-class benchmark performance. See how we’re building intelligent enterprise agents that reason over structured and unstructured data, orchestrate tools and operate with trust. Finally, get a first look at our novel inference and training optimizations and next-generation open-source engine — designed to achieve the elusive trifecta: optimal latency, throughput and cost. Together, these advances enable capable, adaptive and efficient AI systems and applications for enterprise use.