The Generative AI Investment has two “thrust” or focus areas: Cloud-Based AI and High Performance Computing (HPC) Platforms and Applied GenAI. 

Thrust 1: Cloud-Based AI and HPC Platforms

PNNL's unique partnership with Microsoft enables artificial Intelligence research boundaries to be pushed and new technologies to emerge. The following GenAI projects leverage Microsoft's Azure Quantum Elements (AQE) platform for scientific discovery in materials and computational chemistry. 

Carbon Catalysis

• Mining for Catalysts: Chemical Descriptor-Driven Search Over Large Language Model’s Knowledge Space Using Quantum-Chemical Feedback (ChemReasoner)

AI Algorithms and Architectures

• Optimizing Chemical Processing in Real Time for Nuclear Forensics using Multiphysics Modeling (Nuclear Forensics)

Predictive Phenomics

• Quantum-Informed Design for Protein Function (QIPD) 

Thrust 2: Applied Gen AI 

Applying generative AI at the intersection of science, energy, and security can be the catalyst for groundbreaking advancements, providing a compelling opportunity to learn from overlapping challenges and enabling a more holistic understanding of the value (or not) of generative AI. These projects will seek to exceed state of the art (in mission contexts and, if possible, also in AI contexts) through generative AI’s application to one of seven mission domains: 

Grid Modernization and Resilience

• Generative AI Network Model Development Environment for Common Information Model Data (GAINMODE)
• Accelerating Grid Resilience Analysis (Accelysis)
• Intelligent Power System Data Management Platform Based on Fine-Tuned Large Language Models (GenAI_IPSDMP)

Predictive Phenomics and ‘Omics Discovery 

• Generative Protein Design to Secure Energy Self-Sufficiency (Protein Design)
• Protein Function Prediction in Microbial Dark Matter using a Retrieval-Augmented Generation Model Enhanced with Variational Auto-Encoding (RAGEVAE)
• Artificial Intelligence for Advancing Discovery of Post-Translational Modifications in Microbial Proteins (PTMDiscoverer)
• Generating Understanding and Interpretation of Multi-omics Data with an Automated and Generalizable Pipeline (GENRAItOR)

Chemistry and Material Science

• Automatic Generation of New Computational Chemistry Methods in ExaChem (AutoGenCompChem)
• Generative Artificial Intelligence for Predicting Molecular Structures from Measured Signatures in Metabolomics and Chemical Forensics (MoIVisGE)
• Generative Artificial Intelligence for Efficient Prediction of Protein Redox Potentials (RedoxAI)
• High-Performance Gen AI Microscope Data Compression (CompressAI)

Climate and Earth Sciences

• Climate Generator: Learning the Forcing-Response Relationship in Climate Systems (ClimGen)
• Empowering Earth Scientists Through the Use of Generative Artificial Intelligence Tools (empwGenAI)
• Innovative Kilometer-Scale Climate Downscaling using Generative Artificial Intelligence (ICON)

Autonomous Experimentation and Discovery

• Automated Chemical Experiment Design: A Step Toward Self-Driving Labs (AutoLabs)
• Enhancing Laboratory Interactions by Implementing an Artificial Intelligence Assistant in Scientific Research Environments (AI Assistant)
• Learning to Model and Control with Constrained Neural Stochastic Differential Equations (ConFoUnD)
• Autonomous Large Language Model Decision Models (Kranky Kraken)

Nuclear Security

• Information Extraction for Illicit Finance Discovery (FinRED)

Software Engineering

• Large Language Model Pipeline Research (LLM Pipeline)