Atom Grants
Discover
  • Dashboard
  • Search
  • Deep Research
  • Favorites
  • Projects

    Build on Trainium: Agentic, recursive improvement of kernels and frameworks - Fall 2026

    Forecasted

    Build on Trainium: Agentic, recursive improvement of kernels and frameworks - Fall 2026 supports academic researchers developing self-improving AI agents to optimize code and systems on AWS Trainium.

    Browse similar grants
    AI Research Grants 2026AI Safety Research Grants
    OverviewSimilar Grants
    OverviewSimilar Grants

    Funder: Amazon Research Awards

    Due Dates (Anticipated): November 2027 (Submission deadline)

    Funding Amounts: AWS Promotional Credit requests up to $50,000, or up to $250,000 and beyond; one-time awards

    Summary: Supports self-improving AI agents that optimize kernels, frameworks, and training or inference systems on AWS Trainium.

    Key Information: Core compute and research should primarily use AWS Trainium instances; each PI may submit one proposal per call period.


    Description

    Amazon Research Awards’ Build on Trainium call supports academic research on agentic, recursively self-improving systems for kernel and end-to-end optimization. It seeks agents that generate, compile, execute, profile, and repair code using compiler feedback, profiling evidence, and correctness checks. The strongest proposals connect generation, evaluation, and learning, improving performance while reducing optimization time and cost.

    Research priorities include kernel generation and auto-tuning, training and inference workload optimization, performance debugging and explainability, test-harness exploit detection, and benchmarks for correctness and speed. The call also supports supervised fine-tuning and reinforcement learning of open-source models into optimization agents. Learning may improve model weights, retrieved context, tools, or knowledge bases, with an emphasis on reproducible research and open-source contributions.


    Atom

    See the full grant listing

    Sign in to view full eligibility details, sources, similar grants, and AI-powered analysis.