Abhirami Harilal, an Indian American applied scientist specializing in machine learning and statistical modeling for large-scale, real-world systems, is using AI to help unlock one of the universe’s biggest mysteries: dark matter.
Harilal, who earned her BS and MS degrees from Indian Institute of Science Education & Research (IISER), Kolkata, India before getting a PhD in Physics from Carnegie Mellon University, is now helping physicists hunt for new particles.
During her four years at CERN, the European Organization for Nuclear Research, she used machine learning to improve how physicists detect rare particle signatures, which has sharpened the Large Hadron Collider’s search, according to a CMU feature on her.
Along the way, she also upgraded CERN’s detector technology by developing algorithms now deployed inside the Compact Muon Solenoid (CMS) experiment to autonomously identify anomalies more accurately than ever before.
“There are still many fundamental questions about the universe that current physics can’t fully explain,” said Harilal, who graduated from Carnegie Mellon with a PhD in physics. “At the Large Hadron Collider, we’re searching for signs of new particles that could help answer them.”
Only about 5% of the universe is made of matter, which people can see with their own eyes or detect using scientific equipment. The rest of the universe consists largely of dark matter and dark energy, which though unseen affect the movement of galaxies, stars and planets.
Read: Indian American researchers developing tool to spot AI‑generated radiology reports
Researchers at CERN are working to understand what these particles are and how they behave. In 2012, they discovered the Higgs boson, a particle that represents the physical manifestation of a quantum field permeating the universe.
“The Higgs boson takes a very special role in what we know about fundamental particles, and many theories suggest that if new particles exist, they could be connected to the Higgs boson,” Harilal said. “I was trying to see if this Higgs boson would give way or decay into a new particle called an A particle, which could be connected to dark matter or other hidden sectors.”
Harilal used machine learning models to improve how researchers detect the A particle. She used computational modeling to simulate thousands of particle collisions similar to those produced at the LHC. Using these methods, Harilal improved the experiment’s sensitivity to particles with unusual or hard-to-detect signatures, including potentially the A particle.
In addition to conducting her own research, Harilal also improved CERN’s CMS detector using similar machine learning methods to automate its data monitoring system.
Harilal’s machine learning model not only alerts the researchers to unusual activity but also helps in identifying how the data differed from expected results. Her work has made it easier for CERN scientists to diagnose potential issues as they gather data.+“I’m particularly proud about it because this was actually deployed and used during live data taking,” Harilal said.
Harilal is assisting other projects related to the CMS from Pittsburgh. Once she is done with her research, she said she hopes to apply her machine learning skills to other areas in research or industry.
“A big part of my work is recognizing meaningful patterns in large amounts of noisy data, which is also relevant in many other applications,” Harilal said. “Similar ideas apply in areas like medical imaging, financial data, drug discovery. I’m looking forward to finding other opportunities to use these skills.”