Amazon Announces 34 Winners of Build on Trainium Research Grants

Amazon has selected 34 research teams from 30 universities as winners of the Build on Trainium program. This program, totaling $110 million, supports AI research and university education based on AWS Trainium, with participation from Stanford, UC Berkeley, UIUC, UCLA, CMU, and MIT.

This support is the result of a fall 2025 call for proposals focused on Responsible AI. The selected research teams will focus their research on the following areas:

• AI safety and alignment
• Multilingual language models
• Representation engineering
• Sustainability and small language models
• Deep learning models for synthetic data generation

Winning teams will receive Trainium computing credits, access to over 700 Amazon public datasets, and permissions to use AWS AI/ML services and tools. They will also receive mentorship from Amazon researchers and support for Trainium tutorials and hands-on sessions.

Researchers at UIUC will study topology-aware parallelization strategies for mixture-of-experts models with 1 trillion parameters on up to 1,024 Trainium chips. Researchers at the University of Washington will develop a token-efficient optimization framework to implement high-performance, portable LLM inference on Trainium.