Build on Trainium is a $110 million credit program focused on AI research and university education aimed to support the next generation of innovation and development on
__AWS Trainium__This announcement includes awards funded under the Fall 2025 Build on Trainium: Responsible AI call for proposals. Proposals were reviewed for the quality of their scientific content and their potential to impact both the research community and society. This cycle’s focus on Responsible AI invited proposals addressing five priority topics: AI safety and alignment, multi-lingual language models, representation engineering, sustainability and small language models, and deep learning models for synthetic data generation—all leveraging AWS Trainium infrastructure. The recipients have access to more than 700
__Amazon public datasets__Build on Trainium represents AWS's commitment to democratizing AI research through collaborative partnership with academia.
"Build on Trainium gives the next wave of AI researchers powerful, scalable access to Amazon's purpose-built AI chips, so the only limit is their imagination, not their compute budget," said Yida Wang, AWS AI Principal Applied Scientist. "By leveraging the support from Build on Trainium, University of Illinois Urbana-Champaign researchers are studying topology-aware parallelization strategies for large-scale mixture-of-experts models with as many as one trillion parameters on up to 1,024 Trainium chips. At the University of Washington, researchers are developing an inference-optimization framework that raises token efficiency for everyone building on Trainium, with the goal to deliver portable, high-performance LLM inference on Trainium."
| | | |---|---|---| The University of Sydney | FACTOR: Federated Adversarial Co-Training with Textual Gradient for LLM Security and Robustness | | Georgia Institute of Technology | Leveraging Public-Private Mixtures For Differentially Private Synthetic Data Generation | | The University of Queensland | Responsible AI on Trainium: Scalable Detection and Mitigation of Evasive Multimodal Scam Content | | College of William and Mary | DELA: Editable Diffusion Language Models | | University of North Carolina at Chapel Hill | Algorithm-System Co-Design for Efficient Sparse and Quantized LLMs | | University of California Los Angeles | MANSA: Democratizing Voice AI with Efficient Multimodal Foundation Models | | University of California Santa Barbara | Leveraging Public-Private Mixtures For Differentially Private Synthetic Data Generation | | University of California San Diego | Governing Social Bias in AI Image Generation through Value Manifests | | Imperial College London | Towards Deterministic Model Inference | | The Ohio State University | Natively Multimodal and Multilingual Speech-Text Large Language Models | | Institute for Decentralized AI (ADAI) | Safe, Social Pre-training of LLM Agents | | The University of British Columbia | Memorization-Aware Preference Optimization for Machine Unlearning | | New Jersey Institute of Technology | Efficient and Adaptable Language Models via Sub-Model Search | | University of California San Diego | Algorithm-System Co-Design for Efficient Sparse and Quantized LLMs | | The Chinese University of Hong Kong | M3-Align: Scalable Multilevel & Multiobjective Alignment for Multilingual Language Models | | UCL - University College London | Breaking the Multilingual Data Wall: Scaling Synthetic Data for Low-Resource Language Model Pretraining | | University of Illinois at Urbana-Champaign | Cratos: Certified Robustness for Quantization and Pruning-Aware Training and Tuning of Vision Language Models | | Tinoosh Mohsenin | Johns Hopkins University | TRIM-LLM: From Quadratic to Linear Attention and Structured Pruning for Carbon and Cost-Efficient LLM Deployment on Trainium | George Mason University | Leveraging AWS Trainium for Verifiable AI and ML-Assisted Mathematical Reasoning | | Frank Rudzicz | Dalhousie University | Representation Immunization on Trainium: Scalable Noising & Weight-Locking | Anuj Sharma | Iowa State University | Build on Trainium: Physics-Grounded Synthetic Crash Generation for Vulnerable Road Users with Representation Engineering on Video Diffusion and VLMs | Shen Shen | Massachusetts Institute of Technology | Agent Tool-Use Safety Benchmarking with MCP-Specific LoRA Mitigations | Ryan Shi | University of Pittsburgh | Benchmarking and Improving Multilingual LLMs on Real Indic Language Healthcare Dialogues | Naichen Shi | Northwestern University | LLM Hallucination Detection and Mitigation | Jaideep Srivastava | University of Minnesota Twin Cities | Knowledge-Infused Time-Series Pretraining with Safety-by-Knowledge-Checking for Trustworthy Clinical AI | Cheng Tan | Northeastern University | Towards Reliable and Trustworthy LLM Services with ϵ-correctness | Yue Wang | University of Central Florida | Game-Theoretic Frameworks for Responsible AI on Pluralistic Alignment | Yang Wang | University of Illinois at Urbana-Champaign | Safeguarding Youths in Multimodal Generative AI: Toward a Trainium-Powered Framework for Safety and Alignment | Ermin Wei | Northwestern University | Higher Order Based Fast LLM Training Method | Jun Wu | Michigan State University | Bigger Models, Bigger Risks? Investigating the Safety Landscape of LLM Scaling | Xiaokui Xiao | National University of Singapore | Trainium-Accelerated, LLM-Guided Differentially Private Synthesis of Hierarchical Relational Data | Min Xu | Carnegie Mellon University | Language-Grounded Interpretability for ViT and 3D Models | Ziyu Yao | George Mason University | Representation Engineering of LLMs for Secure Code Generation | Junzhe Zhang | Syracuse University | Deconfounding Image Editing for Robust Causal Prediction |