GPU Modeling Research of Smart Modeling and Simulation for HPC (SMASH)

The Department of Energy Office of Science is seeking a contractor for GPU Modeling Research of Smart Modeling and Simulation for HPC (SMASH) at Brookhaven National Laboratory. The pre-solicitation notice requests information from potential contractors capable of developing machine learning-based performance models for GPUs used in scientific computing and machine learning workloads. The primary contract objectives include designing an ML technology to model parallel GPU execution, improving model accuracy through neural architecture search, and demonstrating use cases for ML-based GPU models in design space exploration. Interested contractors must submit a capabilities statement detailing how they plan to meet the Statement of Work requirements and demonstrate their credentials. The response deadline is February 5, 2025, with the pre-solicitation being issued on January 30, 2025.

The contract is not set aside for any specific category of business and falls under the NAICS category of Colleges, Universities, and Professional Schools. The performance period for the project runs from January 1, 2025 to September 30, 2025, with specific milestones including two months for single thread warp execution model development, three months for parallel warp and thread block execution, two months for model optimization, two months for use case exploration, and two months for technical paper writing. Key deliverables include a codebase implementing the ML-based GPU models, due 200 days after receipt of order, and a technical paper describing the implementation and evaluation, due 220 days after receipt of order. While no specific award value is mentioned, the project requires advanced technical capabilities in machine learning, GPU performance modeling, and high-performance computing research.