A high school student working with Caltech researchers has developed a powerful AI-based model that revealed 1.5 million previously unknown objects in space, dramatically expanding the scientific utility of NASA’s retired NEOWISE telescope. The work led to a single-author publication in The Astronomical Journal and could reshape how astronomers study cosmic variability.
From Stargazing To Scientific Breakthrough
The remarkable story begins with Matteo Paz, a local high school student with a long-standing passion for astronomy. After attending public stargazing lectures at Caltech as a child, Paz joined the Planet Finder Academy in 2022, where he began working alongside professional scientists under the mentorship of Davy Kirkpatrick, senior scientist at IPAC.
“I’m so lucky to have met Davy,” Paz said. “I remember the first day I talked to him, I said that I was considering working on a paper to come out of this, which is a much larger goal than six weeks. He didn’t discourage me. He said, ‘OK, so let’s talk about that.’”
Kirkpatrick, inspired by his own high school mentor, saw potential in Paz and helped guide him through an ambitious project involving data from NEOWISE — an infrared telescope launched by NASA to identify near-Earth asteroids. While NEOWISE had scanned the sky for over a decade, its dataset held far more than just asteroid information.
Unlocking A Decade Of Unused Data
Though the telescope was retired, its massive archive — nearly 200 billion rows of observations — contained valuable signals from objects such as quasars, variable stars, and eclipsing binaries. But due to the nature of these time-sensitive phenomena, the data had not been systematically explored for variability.
Kirkpatrick proposed that they examine just a slice of the sky, handpicking new variable stars to demonstrate the potential buried in the archive. But Paz took a different route. With strong foundations in AI and mathematics — developed through advanced coursework in the Pasadena Unified School District’s Math Academy — he set out to automate the process.
“His schoolwork had prepared him to bring a new viewpoint to the challenge,” the article noted. Paz had studied AP Calculus BC in eighth grade and learned to apply machine learning to time-domain data. That skillset allowed him to create an algorithm that could flag candidate objects by detecting subtle variations in infrared brightness over time.
An AI model with Wide Applications
The model, refined over two years of collaboration with astronomers, was eventually able to process the entire NEOWISE dataset. The result: a discovery of 1.5 million previously undetected celestial objects. These variable sources offer insight into the behavior of phenomena like exploding stars, quasars, and binary systems, expanding the scientific reach of NEOWISE far beyond its original mission.
“The model I implemented can be used for other time domain studies in astronomy, and potentially anything else that comes in a temporal format,” Paz explained. “I could see some relevance to (stock market) chart analysis, where the information similarly comes in a time series and periodic components can be critical. You could also study atmospheric effects such as pollution, where the periodic seasons and day-night cycles play huge roles.”
The model has been described as a “Submillisecond Fourier and Wavelet-based Model” — an advanced approach capable of detecting brief or slowly changing events often overlooked by traditional methods.
Mentoring the Next Generation
After his initial work in 2022, Paz returned in 2024 to mentor new high school students in the same academy that launched his research journey. Now a paid employee at Caltech, he works for Kirkpatrick at IPAC, which manages data from missions supported by both NASA and the NSF.
“Every meeting with Davy is 10% work and 90% us just chatting. It’s been super cool just to have someone to talk to about science like that.”
Kirkpatrick, reflecting on his own academic beginnings, added: “If I see their potential, I want to make sure that they are reaching it. I’ll do whatever I can to help them out.”
The full catalog of Paz’s findings is scheduled for release in 2025, opening new doors for astronomical research and exemplifying how AI-driven analysis and early mentorship can yield discoveries on a cosmic scale.
Smart Kid.