Artificial Intelligence and Machine Learning
As the AI Scientist in EBNN, Hendrik Hamann collaborates with domain and computer scientists to advance the state of the art of artificial intelligence and its applicability to improving domain sciences. As a thought leader, he contributes to the overall AI vision and strategy across EBNN and more broadly across Brookhaven Lab, including areas such as electric modernization and geosciences
Hendrik’s research focuses on the intersection of physical and computational sciences, particularly on AI foundation models, machine learning, especially as it relates to advancing scientific discovery.
Previously, Hendrik led five multi-million-dollar Department of Energy-funded projects, four of which he served as the Principal Investigator (PI). These projects focused on developing energy management technologies pioneering the use of AI and machine learning, for example to enhance the resiliency of the electric grid, to advance the accuracy of weather forecasts, to improve data center energy efficiency and to enable clean natural gas extraction.
Environmental Science and Technologies Department
The Environmental Science and Technologies Department is applying Artificial Intelligence and Machine Learning across various areas of atmospheric research. These tools help us analyze complex datasets, improve the design of sensor networks, enhance data quality and calibration, and accelerate development of predictive models.
Autonomous Environmental Laboratories and Smart Sensing
By integrating AI, edge computing, sensor fusion, and advanced radar technologies, we enable real-time sensing of weather phenomena such as thunderstorms, wind patterns, and aerosol plumes. These capabilities are used to support field campaigns and targeted studies of atmospheric variability in regions important for energy systems, national security operations, and air quality monitoring.
Cloud and Precipitation Modeling
We are developing AI-based emulators to represent the behavior of clouds and aerosol particles, supporting the creation of “digital twins” for experimental cloud chambers and hybrid modeling of cloud and precipitation processes. These efforts help improve predictions of rainfall and extreme weather, which are essential for water resource management, agriculture, and public safety.
Pattern Classification in Complex Datasets
Unsupervised and semi-supervised ML approaches help us identify patterns in large and diverse atmospheric datasets. For instance, we have used these techniques to classify snowflake and aerosol particle properties from data collected by the DOE Atmospheric Radiation Measurement (ARM) facility. This work improves our ability to model severe weather, wildfire impacts, and precipitation. Ongoing efforts are also focused on classifying regional weather patterns and studying their effects on local aerosol and cloud behavior.
Causal Discovery in 4D Datasets
We apply AI in combination with causal modeling to analyze feedbacks between aerosols, clouds, and precipitation. These studies help improve our understanding of how natural and anthropogenic particles influence rainfall, drought, and energy flow through the atmosphere. This knowledge supports applications in agriculture, water planning, and national defense.
Sampling Network Optimization and Spatial Variability Reconstruction
We use AI-driven diffusion models to optimize the placement and coordination of sensors in complex environments like cities and forests. These methods are applied in Phoenix as part of a DOE Integrated Field Laboratory, and in Alabama at the Bankhead National Forest. In parallel, we use similar techniques to reconstruct 3D environmental fields from sparse data—for example, mapping clouds or aerosol plumes using drones and distributed sensor networks.
We leverage large language models (LLMs) to support day-to-day scientific activities, including the generation of plain-language research summaries, identification of cross-disciplinary research opportunities, acceleration of software development, and creation of scientific diagrams and figures.
Nonproliferation and National Security Department
The Nonproliferation and National Security Department (NN) is investigating applications of artificial intelligence and machine learning (AI/ML) in nonproliferation and national security mission space. NN develops conventional ML methods and deep learning approaches to improve deployed technologies in this domain as well as exploring novel applications of advanced foundation models to improve the effectiveness and efficiency of operation. In addition, NN is also working on AI/ML with emerging technologies like robotics to build cross-cutting tools to fulfill the needs of operators and address challenges in nonproliferation and national security.
Nuclear Science and Technology Department
The Nuclear Science and Technology Department uses AI/ML in experimental, theoretical, engineering, and data efforts:
- Automate data analysis for fast throughput, including signal identification, background subtraction, pattern and anomaly detection.
- Parameter optimization in theoretical models, leading to more accurate predictions where no data exists, or data are discrepant.
- Design, optimization, safety, and safeguard analysis of nuclear reactors.
- Training of language models to extract data and generate metadata from nuclear physics literature.
- Development of neural networks to generate correlations, identify outliers, and to generate recommended data including realistic uncertainties.
Biology
The Biology Department is dedicated to advancing the U.S. Department of Energy's mission to foster a sustainable bioeconomy in the United States, emphasizing the production of biofuels and bioproducts. The department leverages artificial intelligence and machine learning (AI/ML) to tackle diverse biological challenges, including protein functional prediction, structural and imaging data analysis, enzyme design, and plant host-pathogen interactions. By integrating AI/ML with experimental approaches, our researchers accelerate their plant biology research and development activities.
