Amazon Awards 63 Researchers for Groundbreaking AI and Health Innovations
Amazon has announced the recipients of its Spring 2025 Amazon Research Awards (ARA), granting unrestricted funds and AWS credits to 63 academic researchers across 41 universities in 8 countries. The awards support innovative projects spanning AI for Information Security, Amazon Ads, Agentic AI, and advancements utilizing AWS Trainium, alongside the "Think Big" initiative.
Key Takeaways
- 63 researchers from 41 institutions globally have been recognized.
- Awards focus on critical areas like AI security, advertising, agentic AI, and health.
- Recipients gain access to Amazon's datasets and AWS AI/ML services.
Driving Innovation Across Disciplines
The Amazon Research Awards program aims to foster impactful research by providing crucial resources to academic institutions. This cycle's awards are distributed across five key areas: AI for Information Security, Amazon Ads, AWS AI: Agentic AI, Build on Trainium, and Think Big. Proposals were evaluated based on their scientific merit and potential societal impact.
Advancing Healthcare Through AI
Christine Silvers, AWS Principal Healthcare Advisor, highlighted the transformative potential of ARA-funded research in healthcare. Projects include revolutionizing structural biology tools to accelerate drug discovery, predicting stroke causes for timely treatment, and interpreting digital phenotyping data to enhance mental health services. "The potential for improving healthcare amongst all of the spring 2025 plus past and future awardees is staggering and inspiring,” Silvers stated.
Democratizing AI Research with AWS Trainium
Yida Wang, AWS AI Principal Applied Scientist, emphasized how the "Build on Trainium" program addresses the challenge of accessing powerful and affordable AI infrastructure for academic researchers. Universities like UC Berkeley, Stanford, and MIT are leveraging AWS Trainium to achieve significant breakthroughs, such as improving state-of-the-art performance in areas like FlashAttention and accelerating the training of 3D medical imaging models.
Spring 2025 Award Recipients by Research Area
AI for Information Security
| Recipient | University | Research Title |
|---|---|---|
| Christopher Fletcher | University Of California, Berkeley | Design and Verification of High-Assurance Key Management Services for Stateful Confidential Computing |
| Zhou Li | University Of California, Irvine | Precise and Analyst-friendly Attack Provenance on Audit Logs with LLM |
| Yu Meng | University of Virginia | Weakly-Supervised RLHF: Modeling Ambiguity and Uncertainty in Human Preferences |
| Jelena Mirkovic | University of Southern California | Safe and Secure API Discovery for Agentic AI |
| Aanjhan Ranganathan | Northeastern University | Understanding How LLMs Hack: Interpretable Vulnerability Detection and Remediation |
| Sanjit Seshia | University Of California, Berkeley | Design and Verification of High-Assurance Key Management Services for Stateful Confidential Computing |
| Alexey Tregubov | University of Southern California | Safe and Secure API Discovery for Agentic AI |
| Ziming Zhao | Northeastern University | Understanding How LLMs Hack: Interpretable Vulnerability Detection and Remediation |
Amazon Ads
| Recipient | University | Research Title |
|---|---|---|
| Xiaojing Liao | University of Illinois at Urbana–Champaign | Adversarial Misuse of Large Language Models in Digital Advertising: Benchmarking and Mitigation |
| Tianhao Wang | University of Virginia | Adversarial Misuse of Large Language Models in Digital Advertising: Benchmarking and Mitigation |
AWS Agentic AI
| Recipient | University | Research Title |
|---|---|---|
| Faez Ahmed | Massachusetts Institute of Technology | AutoDA-Sim: A Multi-Agent Framework for Safe, Aesthetic, and Aerodynamic Vehicle Design |
| Fabio Anza | University of Maryland, Baltimore County | Physics Co-Pilot: An LLM-Orchestrated Scientific Assistant for Physics Research |
| Andrea Bajcsy | Carnegie Mellon University | Fine Grained Planning Evaluation for VLM Web Agents |
| Niranjan Balasubramanian | Stony Brook University | Efficient and Effective Long-Horizon Reasoning for Interactive LLM Agents |
| Andreea Bobu | Massachusetts Institute of Technology | Contextual Harm Mitigation and Automated Backtracking in Computer Use Agents |
| Joseph Campbell | Purdue University, West Lafayette | Open-World Probabilistic Theory of Mind |
| Cong Chen | Dartmouth College | Empowering Power Systems and Market Operations with Behavioral Generative Agents |
| Chunyang Chen | Technical University of Munich | Functional Bug-Aware Software Testing via Intelligent Computer Use Agents |
| Shay Cohen | University of Edinburgh | Diffusion-inspired chain-of-thought self-revision |
| Fernando De la Torre | Carnegie Mellon University | Fine Grained Planning Evaluation for VLM Web Agents |
| Sidong Feng | Monash University | Functional Bug-Aware Software Testing via Intelligent Computer Use Agents |
| James Fogarty | University of Washington, Seattle | Leveraging Multiple Representations in Multi-Agent Mobile App Interface Understanding and Task Execution |
| Surbhi Goel | University of Pennsylvania | Efficient and Safe Protocols for Collaborative Agentic AI |
| Nika Haghtalab | University of California, Berkeley | Multi-Agent AI Alignment |
| Irwin King | The Chinese University of Hong Kong | WebAGI: VLM-Driven Framework for Robust Web Automation and Planning in Agentic AI |
| Emma Lejeune | Boston University | Formidable yet Solvable: Scientific Computing Tasks for Agentic AI |
| Bang Liu | University of Montreal | Foundation Agents and Protocol for Collaborative Agentic AI |
| Harsha Madhyastha | University of Southern California | Improving the Efficiency of Web Agents |
| Michael Macy | Cornell University | Artificial Collective Intelligence: The Structure and Dynamics of LLM Communities |
| Radu Marculescu | University of Texas at Austin | Collaborative Continual Learning in Multimodal Multi-Agent Systems |
| Lianhui Qin | University of California, San Diego | ReaL-Agent: A Retrieval-and-Reasoning Agent for Deep, Cross-Modality Retrieval |
| Mahnam Saeednia | Delft University of Technology | Heterogeneous Multi-Agent Collaboration For Built-in Resilience |
| Maarten Sap | Carnegie Mellon University | OpenAgentSafety: Measuring and Mitigating Safety Harms of LLM-based AI Agent Interactions |
| Vitaly Shmatikov | Cornell University | Contextual Security for Multi-Agent Systems |
| Haim Sompolinsky | Harvard University | Lifelong learning in agentic AI through gated memory modules |
| John Torous | Harvard University | Interpreting Digital Phenotyping Data with LLM-Based Agentic Assistants for Mental Health Services |
| Jindong Wang | College of William & Mary | Structure Matters: Task-Optimized Topologies for LLM Agents |
| Xiaolong Wang | University of California, San Diego | Agentic World Representation |
| Zhi-Li Zhang | University of Minnesota, Twin Cities | NetGenius: Agentic AI for Next-Generation Wireless Network Autonomous Configurations and Intelligent Operations |
| Jiawei Zhou | Stony Brook University | Efficient and effective long-horizon reasoning for interactive LLM agents |
Build on Trainium
| Recipient | University | Research Title |
|---|---|---|
| Saikat Dutta | Cornell University | VERA: Automated Testing for Improving the Reliability of Neuron Compiler Toolchain |
| Kuan Fang | Cornell University | Fast Adaptation of Multi-Modal Foundation Models for Robotic Perception and Control |
| Shizhong Han | Lieber Institute for Brain Development | Optimizing and scaling pretraining and preference-based fine-tuning of Large Chemical Models |
| Sitao Huang | University of California, Irvine | Automatic Kernel Synthesis and Tuning for AWS Trainium via Profile-Guided Graph Topology Optimization |
| Wataru Kameyama | Waseda University | Accelerating Vision-Language Autonomous Driving with AWS Trainium |
| Dong Li | University of California, Merced | Efficient Sparse Training with Adaptive Expert Parallelism on AWS Trainium |
| Xiaoxiao Li | University of British Columbia | Efficient MoE LLMs via Pruning and Matryoshka Quantization on AWS Trainium |
| Jiang Liu | Waseda University | Accelerating Vision-Language Autonomous Driving with AWS Trainium |
| Xiaoyi Lu | University of California, Merced | Accelerating Large Language and Reasoning Model Workloads with AWS Trainium |
| Satoshi Masuda | Tokyo City University | LLM for Software Modeling Brain in Multi Language |
| Andrew McCallum | University of Massachusetts, Amherst | Overcoming Fundamental Reasoning Limitations of LLMs by Always Thinking before Writing |
| Xupeng Miao | Purdue University, West Lafayette | Towards Communication-Efficient Distributed Training of Large Foundation Models by Dataflow-aware Optimizations |
| Michael Nagle | Lieber Institute for Brain Development | Optimizing and scaling pretraining and preference-based fine-tuning of Large Chemical Models |
| Jean-Christophe Nebel | Kingston University London | Efficient Architectures for Genomic Variant Interpretation: Language Models for Non-Coding DNA Variant Analysis |
| Farzana Rahman | Kingston University London | Efficient Architectures for Genomic Variant Interpretation: Language Models for Non-Coding DNA Variant Analysis |
| Rohan Sachdeva | University of California, Berkeley | Learning Host–Microbial Genetic Element Interactions with Genomic Language Models |
| Yanning Shen | University of California, Irvine | Automatic Kernel Synthesis and Tuning for AWS Trainium via Profile-Guided Graph Topology Optimization |
| Yun Song | University of California, Berkeley | Learning Host–Microbial Genetic Element Interactions with Genomic Language Models |
| Hoa Vo | Indiana University Bloomington | AI-Powered Travel Pattern Detection in VR for Occupant Behavior Analysis Using AWS Trainium |
| Minjia Zhang | University of Illinois Urbana-Champaign | Trainium-native MoE: Developing kernel and system optimizations for efficient and scalable MoE training |
Think Big
| Recipient | University | Research Title |
|---|---|---|
| Tianlong Chen | University of North Carolina at Chapel Hill | Leveraging Molecular Dynamics to Empower Protein AI Models |
| William H. Lee | Yale School of Medicine | AI-powered prediction of ischemic stroke etiologies using multi-modal data |
| Piotr Sliz | Harvard Medical School | SBCloud – A Transformative Model for Scalable Structural Biology Research |