Amazon Funds 34 Cutting-Edge AI Research Projects
Next-gen AI innovations to be developed using AWS Trainium chips.
1$110 Million AI Research Initiative
Amazon's Build on Trainium program has launched a $110 million credit initiative to support AI research and university education. The program provides compute credits for academic teams developing novel model architectures, ML libraries, and optimizations on AWS Trainium chips. Starting Fall 2025, it focuses on 'Responsible AI' proposals, evaluated based on scientific merit and societal impact potential. Selected projects leverage Trainium infrastructure for groundbreaking work.
2Five Pillars of Responsible AI
This cycle prioritizes projects addressing AI safety/alignment, multilingual models, representation engineering, sustainable small language models, and synthetic data generation via deep learning. Researchers gain access to 700+ Amazon public datasets and AWS AI/ML services through promotional credits. Each team receives dedicated Amazon research support, tutorials, and hands-on training with Trainium resources.
3Groundbreaking University Projects
University of Illinois researchers are developing topology-aware parallelization for trillion-parameter mixture-of-experts models scalable to 1,024 Trainium chips. At University of Washington, teams are creating an inference-optimization framework to enhance token efficiency for portable, high-performance LLM inference on Trainium. Both exemplify the program's ambition to push AI boundaries.
4Global Research Collaboration
The University of Sydney's Wei Bao is pioneering federated learning for LLM security, while Georgia Tech's Viveck Cadambe works on differentially private synthetic data. These projects span three continents, demonstrating the program's international reach. Other recipients include institutions from Australia, UK, Singapore, and Canada, fostering cross-border AI innovation.
5Industry-Academia Partnership Model
Amazon assigns dedicated research contacts to each team, providing technical consultation. Beyond Trainium access, researchers benefit from AWS tutorials and workshops. AWS Principal Applied Scientist Yida Wang states: 'Build on Trainium removes compute barriers, letting researchers' imagination be their only limit.' The program bridges cutting-edge academic research with industrial-scale cloud infrastructure.
6Transformative Academic Work
UNC's Tianlong Chen is co-designing algorithms for efficient sparse/quantized LLMs. UCLA's Saadia Gabriel aims to democratize voice AI via multimodal foundation models. Imperial College's Marios Kogias pursues deterministic model inference methods. These projects tackle fundamental challenges in AI scalability, accessibility, and reliability using Trainium's specialized architecture.
7Ethics-Focused AI Advancements
UC San Diego's Haojian Jin develops frameworks to govern social bias in AI image generation. University of Minnesota's Jaideep Srivastava creates knowledge-infused time-series models for trustworthy clinical AI. Such initiatives align with growing demands for ethical AI development, addressing bias mitigation and accountability in mission-critical applications.
8Redefining AI Research Frontiers
By eliminating computational constraints, Build on Trainium accelerates AI innovation across 34 university projects. From multilingual model scaling to verifiable AI reasoning, these initiatives promise breakthroughs in foundational AI challenges. Amazon's investment not only advances academic research but also stress-tests AWS's AI infrastructure capabilities, driving the entire field forward.
Original article: Amazon Science — Amazon 34 AI Araştırma Projesine Destek Verdi