A University of Oregon-led research team has received a $662,000 phase one grant from the U.S. Department of Energy (DOE) as part of the Genesis Mission, a national initiative to bring together universities, DOE national labs, and industry together to integrate artificial intelligence (AI) into science, energy, and national security. The Department of Energy received more than 5,000 applications and awarded funding to 278 projects; these highly selective awards were announced in Washington, D.C. on July 22, 2026.
Joined by collaborators at Brown University and Brookhaven National Laboratory, physicist and professor Stephanie Majewski is leading a project that will use AI to save time and computing costs for experiments at CERN’s Large Hadron Collider (LHC), the world’s most famous particle accelerator—a massive underground experiment that smashes together the smallest particles that make up matter with a level of energy near that of the Big Bang.
“Particle physics at the LHC, a flagship DOE program at the energy frontier, generates some of the most complex datasets in science,” Majewski said. “MANGO shows how the Genesis Mission’s investment in AI-driven scientific computing translates directly into faster, cheaper insight from that data.”
MANGO, an abbreviation for Monte Carlo Acceleration via Normalizing Flows using GPU Optimization, rethinks Monte Carlo event generation—a process by which particle physicists simulate what they think will happen when particles collide so they can compare their predictions against real experimental data, then use that information to fine tune future experiments. Each collision at the LHC produces terabytes of raw data that scientists must parse.
Monte Carlo event generation is a bottleneck standing between raw collider data and physics discovery. MANGO doesn’t just speed that bottleneck up, it re-thinks the entire generation process to produce simulations that are both dramatically faster and more accurate, a combination that hasn't existed before.
“Faster Monte Carlo event generation is really about how efficiently the country’s multi-billion-dollar investment in fundamental science gets turned into discovery,” she said.
Additionally, the project’s optimization process, if successful, can be implemented in computing farms that use both graphics processing units (GPUs, which process images particularly well) and central processing units (CPUs, the “brain” of a computer) and doesn’t require special hardware. This optimization could help experiments at CERN and the work of particle theorists. Majewski is one of several UO physicists with long-time affiliation with CERN and member of the UO’s Institute for Fundamental Science. She is currently the deputy project manager for an upgrade to prep CERN’s detectors for the accelerator upgrade called the High Luminosity LHC that seeks to answer the question: What is dark matter?
“This project sits at the intersection of machine learning and particle physics, and which is where the UO ATLAS group specializes,” Majewski said. “My group has gained essential expertise investing in high-performance, GPU-accelerated systems that must run at the scale and speed the LHC demands. MANGO is a natural extension of that expertise into simulation.”
One of MANGO’s main goals is to make the process applicable to different types of hardware so that it’s independent of a particular GPU brand, which are manufactured by multiple companies like NVIDIA and AMD.
“Every dollar we save on computing is a dollar that goes further toward finding new physics and toward keeping the U.S. at the forefront of a field it helped invent,” she said. “Projects like MANGO also train the next generation of scientists at the intersection of AI and physics, providing students and researchers with skills the broader economy needs, whether they stay in fundamental research or move into industry.”