Microsoft Build x Anyscale

Details coming soon!

Multimodal AI has unlocked 90% of the world's data: video, audio, PDFs, and sensor streams. But multimodal pipelines interleave CPU-bound preprocessing with GPU-bound inference, and traditional batch architectures leave GPUs idle over 50% of the time.

In this session you will learn:

• Why multimodal data pipelines break traditional batch engines
• How to avoid CPU → GPU I/O that drives +50% GPU idle time
• How to build a streaming CPU+GPU pipeline with Ray Data

You'll leave with a framework for scaling multimodal pipelines and keeping your GPUs busy with Ray and Anyscale.