From Silicon to Solution: Building a World-Class Inference Engine for the AI Builder

​Join GMI Cloud and Yujing Qian (Head of Engineering) for a deep dive into how modern inference systems are evolving in the era of next-gen GPUs.

​As new architectures like Blackwell push the limits of compute, many production systems face unexpected bottlenecks in latency, cost, and scaling. This session explores what actually changes at the system level — and how to rethink your inference stack for real-world workloads.

​What You’ll Learn

​• Why traditional inference assumptions break on next-gen GPU architectures
• How to redesign batching, scheduling, and concurrency strategies
• Key architectural shifts for better latency and cost efficiency at scale
• What real production traffic reveals about inference system behavior

​Why Attend

​If you're building or scaling AI inference systems, this session will give you a clearer framework for moving from raw compute power → production performance.

​⚡ Join us live at Booth #142.