Inference Time Compute Optimization Workshop

​Inference-time compute optimization is about making AI systems cheaper, faster, and more scalable in production. While most of the industry focuses on training bigger models, the real constraint for startups is what happens after deployment, every prompt, every token, every millisecond adds up.

​In this workshop, we’ll break down how to design systems that minimize inference cost without sacrificing performance.

​We’ll cover practical strategies like model routing between small and large models, quantization, batching, caching, distillation, and when to fine-tune versus prompt engineer. This session is built for founders and engineers shipping AI products who want to understand the economics behind their architecture decisions and build systems that scale sustainably.