In a discovery that sounds more like science fiction than scientific fact, a local high school student has illuminated the universe with 1.5 million previously undetected cosmic objects—all thanks to a cutting-edge artificial intelligence algorithm he developed himself. Matteo (Matthew) Paz’s journey from grade-school stargazer to published astrophysics researcher is not only remarkable, it’s a testament to the power of curiosity, mentorship, and the untapped potential of youth in the age of AI.
His breakthrough has been formally recognized in The Astronomical Journal, where Paz published a rare single-author scientific paper—a feat seldom achieved by even seasoned graduate students, let alone a teenager still finishing high school. But this is no ordinary student, and this is no ordinary story.
From Stargazing Lectures to Star Catalogs
Paz’s interest in the cosmos began early, sparked by evenings spent with his mother at public stargazing lectures hosted by Caltech. These events weren’t just recreational; they planted the seeds of a lifelong passion. Years later, in the summer of 2022, that passion would take a dramatic turn when he joined Caltech’s Planet Finder Academy—a program designed to immerse high school students in the frontier of astronomical research. It was there that he met astronomer and senior scientist Davy Kirkpatrick, who would become his mentor and advocate.
“I’m so lucky to have met Davy,” Paz reflects. “He allowed an unbridled learning experience. I think that’s why I’ve grown so much as a scientist.”
Kirkpatrick, a veteran astronomer whose own journey began in a small Tennessee town, was inspired by a mentor of his own—ninth-grade teacher Marilyn Morrison—who once told him he had potential. Now, he was paying it forward, giving Paz not just the freedom to explore, but the tools and guidance to aim higher than anyone expected.
The Sleeping Giant in the Data
At the heart of this story lies NEOWISE, NASA’s Near-Earth Object Wide-field Infrared Survey Explorer. Originally tasked with scanning the skies for potentially hazardous asteroids, the infrared telescope quietly collected over a decade’s worth of data—nearly 200 billion data points in total. Though the telescope retired, its treasure trove of information remained largely unexplored, particularly when it came to variable celestial objects: stars that dimmed, flared, pulsed, or flickered with cosmic rhythm.
Most astronomers focused on NEOWISE’s asteroid-tracking mission, but Kirkpatrick saw an untapped goldmine.
“The idea for the summer,” he said, “was to take a little piece of the sky and see if we could find some variable stars. Then we could say, ‘Here’s some new stuff we discovered by hand; just imagine what the potential is in the dataset.’”
But Paz had no intention of combing through billions of data points manually. His tool of choice was artificial intelligence—specifically, machine learning.
Building an AI to Decode the Universe
Paz’s background in theoretical computer science and mathematics gave him the perfect foundation to tackle the problem. Enrolled in Pasadena Unified School District’s elite Math Academy—where students complete AP Calculus BC by eighth grade—he had already been exposed to high-level thinking and algorithmic design.
He recognized immediately that NEOWISE’s dataset, vast and structured, was a dream scenario for training a neural network.
With Kirkpatrick’s help, Paz learned the necessary astrophysics and began building a custom AI model designed to detect time-domain anomalies—objects that varied in brightness over time. The initial prototype showed promise. In a mere six weeks, he had created a working machine-learning model capable of identifying patterns even veteran scientists had missed.
Throughout the process, their working relationship grew into something more than mentorship—it became a collaboration of equals.
“Every meeting with Davy is 10% work and 90% us just chatting,” Paz said with a grin. “It’s been super cool just to have someone to talk to about science like that.”
A Village of Experts, A Universe of Insight
To sharpen his algorithm, Paz connected with some of Caltech’s most accomplished astronomers: Shoubaneh Hemmati, Daniel Masters, Ashish Mahabal, and Matthew Graham. Each brought deep expertise in astronomical machine learning and the behavior of transient celestial phenomena.
Their feedback was vital. It helped Paz understand the limitations of NEOWISE’s observational rhythm—it captured long timescales well but struggled with fast-fading or ultra-gradual light variations. Even so, the refined model he developed was powerful enough to begin producing staggering results.
1.5 Million Discoveries—and Counting
By 2024, Paz’s AI had matured into a finely tuned instrument of discovery. It analyzed the raw NEOWISE dataset and flagged 1.5 million candidate objects that exhibited significant brightness variability—quasars, variable stars, pulsating phenomena, and likely supernovae, among others.
What makes these findings extraordinary is not just the volume, but the diversity. These are the kinds of objects that can teach us about stellar evolution, galaxy dynamics, and the birth and death of stars. Some might even hint at phenomena we’ve never seen before.
The duo plans to publish a complete catalog of the findings in 2025, opening the door for the global astronomical community to mine and explore this new cosmic map.
Beyond Astronomy: The Algorithm’s Expanding Universe
Perhaps even more intriguing than the discoveries themselves is the AI engine behind them.
“The model I implemented can be used for other time domain studies in astronomy,” Paz explains. “But I could also see some relevance to stock market analysis, where the information similarly comes in a time series and periodic components can be critical. You could study atmospheric effects like pollution, where seasonal and diurnal cycles play a huge role.”
In other words, this algorithm is a Swiss Army knife for complex, time-based data—a tool with potential far beyond the stars.
From Intern to Caltech Employee
Today, while most high schoolers are worried about college applications, Paz holds a paid position at Caltech’s IPAC (Infrared Processing and Analysis Center). There, he continues working under Kirkpatrick, contributing to real NASA missions, and mentoring younger students in the very program that launched his own journey.
It’s rare for a teenager to be both a published author in a leading scientific journal and a NASA data scientist. But for Matteo Paz, this is just the beginning.
A Glimpse of the Future
Paz’s work is a powerful reminder of what happens when curiosity is nurtured, mentorship is strong, and a young mind is given the freedom to explore. His story is inspiring not only for what he’s accomplished, but for the promise it represents—that the next big scientific breakthrough might come from someone who hasn’t even finished high school.
In a universe that still holds countless mysteries, it’s a comforting and thrilling thought that our next generation of explorers is already pointing their algorithms toward the stars. And perhaps, like Matteo Paz, they’ll find that the cosmos has been waiting for them to look a little closer—and to think a little differently.
Reference: Matthew Paz, A Submillisecond Fourier and Wavelet-based Model to Extract Variable Candidates from the NEOWISE Single-exposure Database, The Astronomical Journal (2024). DOI: 10.3847/1538-3881/ad7fe6