Autodesk's ML Inference Optimization - Leveraging AWS AI Chips

Discover how Autodesk partnered with AWS Generative AI Innovation Center to optimize spatial GenAI models using Inferentia and Trainium chips in this 26-minute conference talk from AWS re:Invent 2025. Learn about the collaboration that achieved a 45% reduction in latency and 62% improvement in price-per-inference across workflows generating 3D objects and CAD sketch constraints. Explore Autodesk's migration journey and gain insights into strategies for implementing custom kernels via the Neuron Kernel Interface. Understand the sustainability benefits and economic advantages of adopting custom silicon, including reduced carbon emissions and optimized infrastructure costs for machine learning inference workloads.