PNNL

The co-design Center for Artificial Intelligence-focused Architectures and Algorithms (ARIAA) promotes the development of core technologies important for the application of artificial intelligence (AI) to Department of Energy (DOE) mission priorities. The center will explore how AI and machine learning can support four key areas: the power grid, cybersecurity, graph analytics, and computational chemistry.

ARIAA four main high-level objectives:

• Co-design novel architectures, algorithms, and compiler and programming abstractions to enable AI-based DOE applications and support sparse, explainable, and domain-informed AI models.
• Understand how AI-focused architectures can accelerate traditional, emerging, and AI DOE workloads.
• Understand and design AI accelerators for future explainable and domain-aware AI methods.
• Increase confidence in AI/machine learning methods for scientific workloads.

The center brings together AI researchers from Pacific Northwest National Laboratory (PNNL), Sandia National Laboratories, and the Georgia Institute of Technology as a collaborative space to simulate and evaluate AI hardware accelerators for future supercomputers. ARIAA researchers focus on holistic hardware/software “co-design,” which balances the capabilities of hardware and software to optimize computer algorithms and architectures, to make improvements when AI and machine learning methods replace or augment more traditional computation methods. ARIAA co-design methodology works on the hardware/software stack level and develops novel applications, runtime, compiler, and hardware designs. Ultimately, this research will help design future supercomputers equipped to solve the nation’s most challenging problems in science, engineering, health, and energy.

ARIAA is supported by the DOE Advanced Scientific Computing Research program and works in close collaboration with the DOE’s Artificial Intelligence and Technology Office. ARIAA is directed by Pacific Northwest National Laboratory senior research scientist and team leader Roberto Gioiosa.