Pacific Northwest National Laboratory - PNNL

Pacific Northwest National Laboratory (PNNL) specializes in research in the fields of chemistry, Earth sciences, biology, and data science for national security missions, nuclear materials stewardship, the nuclear fuel life cycle, on-proliferation missions, energy production, and more. In addition, PNNL is engaged in expanding the beneficial use of nuclear materials, such as nuclear process engineering, radiomaterials characterization, separation, and processing. PNNL also provides support to the Hanford Site cleanup and river corridor protection missions.

PNNL's main objectives are to enable innovations that advance sustainable energy efforts through decarbonization and energy storage and improve national security through the study of nuclear materials and threat analyses. PNNL collaborates with academic institutions in fundamental research and organizations in related industries to transition technologies to market.

In the fiscal year 2020, PNNL employed 4997 scientists, engineers, and professional staff; reached

a total of 2,886 US and foreign owned patents since 1965; published 1,280 peer-reviewed articles; and completed 340 inventions. The laboratory also achieved $1.1B in annual spending and $487M total payroll. Since 1969, PNNL has accumulated 211 FLC and R&D 100 awards and has helped found 198 companies.

PNNL’s computing research ranges from basic to applied. It encompasses data and computational engineering, high-performance computing, applied mathematics, and semantic and human language technologies. PNNL's scientists and engineers develop and apply advanced theories, methods, and models to domains in machine learning, data and computing architectures, and systems integration, as well as software and application development.

Together with the University of Washington, PNNL established the Northwest Institute for Advance Computing (NIAC). It is a physical and virtual center built with the aim of increasing the impact of computing for transformational discoveries. PNNL is also a participating lab in the Exascale Computing Project, a collaborative effort of the US Department of Energy’s (DOE’s) Office of Science and the National Nuclear Security Administration to provide national security and scientific solutions to the DOE.

PNNL has developed an open-source platform that uses deep reinforcement learning to help power system operators create emergency control protocols for electric grid applications. In addition, as part of the Physics-Informed Learning Machines for Multiscale and Multiphysics Problems (PhILMs) Center, PNNL develops physics-informed machine learning techniques.

PNNL’s content intelligence research is oriented towards the development of new AI models to explain and predict social systems and behaviors in light of national security. The laboratory's expertise in descriptive, predictive, and prescriptive analytics includes the detection and attribution of disinformation and forecasting real-world events, such as influenza outbreaks and the price of cryptocurrencies. PNNL employs interactive tools such as CrossCheck, ESTEEM and ErrFilter to develop AI models and advance understanding of large volumes of dynamic, multilingual, and diverse real-world data.

PNNL combines data engineering (development of data architectures and pipelines, data collection, and validation solutions) with artificial intelligence to provide solutions to critical mission spaces.

Sandia National Laboratories and Georgia Institute of Technology were commissioned by the US Department of Energy to establish a research center for Artificial Intelligence-focused Architectures and Algorithms (ARIAA). The ARIAA team is focused on researching the application of artificial intelligence to address the US Department of Energy in areas such as cybersecurity and graph analytics. PNNL researchers integrate high-performance computing, deep learning, and graph analytics to accelerate scientific discovery.

PNNL carries out almost all research on few-shot learning (FSL) exclusively on images, although the laboratory has conducted successful experiments in other data types, including text, audio, and video. According to PNNL, the capability to work with various data types has expanded its AI capabilities beyond traditional, publicly available image datasets and allows researchers to efficiently build machine learning models using small amounts of user-classified training examples. The web application Sharkzor, for instance, combines machine learning techniques with human interactions to enable classification using five to ten images, as opposed to hundreds or thousands that are required for traditional deep learning methods.

PNNL approaches scientific problems on the basis of mathematical theory through an iterative analysis process involving exploratory data analysis; formal experimentation; and the development and implementation of multiscale, multiphysics, and machine learning algorithms. The laboratory designs resource utilization strategies and develops artificial intelligence systems that can sense application environments and adapt algorithms to dynamic systems. These solutions are applied in the deployment of limited resources, supply chain implementation, and analysis of operations to manage risks and achieve operational objectives.

In addition, PNNL researches, designs, implements, and validates software, tools, and statistical package products for use in bioremediation multiscale modeling, industrial engineering, climate research, and human behavior modeling. The laboratory's application of mathematics and statistics aims to provide insight into physical, chemical, and biological principles using computational modeling, experimentation, and data evaluation.

The PNNL software engineering department develops software systems for a variety of applications, including software architecture, technology assessment, requirements analysis, software quality, user experience, software testing, and documentation. PNNL's deployed systems include international nuclear materials safeguards and security and radiation detection systems.