Funder: Amazon Research Awards
Due Dates: November 4, 2026, 11:59 PM Pacific Time (Submission) | February 2027 (Decision letters)
Funding Amounts: AWS Promotional Credits: up to $50,000 or up to $250,000 and beyond; one-time awards
Summary: Supports self-improving AI agents that optimize kernels and end-to-end training and inference workloads on AWS Trainium.
Key Information: Proposals should primarily use AWS Trainium instances for core compute and research.
Description
Amazon Research Awards (ARA) supports research into agentic, recursively self-improving systems for kernel and end-to-end workload optimization. This call is part of Build on Trainium, a $110 million credit initiative supporting AI research and university education. It seeks systems that generate, run, profile, and repair code using compiler feedback, profiling evidence, and correctness checks, then improve their own models, context, tools, or knowledge bases over time.
Research priorities include kernel generation and auto-tuning, training and inference system optimization, efficient optimization loops, performance debugging and explainability, correctness benchmarks, and reward-hacking detection. The call also supports supervised fine-tuning and reinforcement learning of open-source models for optimization. The strongest proposals connect generation, evaluation, and learning, with particular interest in substantial reductions in optimization time and cost and reproducible artifacts the community can extend.
