Market Data 
- High schooler Matteo Paz won $250,000 for discovering 1.5 million new space objects with AI.
- Paz built an AI algorithm to search data from a NASA space telescope for objects like black holes.
- His discoveries could help solve one of the universe’s biggest mysteries.
When Matteo Paz scored a high school internship at the California Institute of Technology, the scientists there gave him the daunting task of manually sorting reams of data from a NASA mission.
It was “classic intern work,” Paz, an 18-year-old from Pasadena, California, told Business Insider. “The very menial, tedious, dirty tasks that require a lot of time.”
Instead of manually sifting through the data, Paz built an AI algorithm to do it for him. Ultimately, he discovered 1.5 million new objects in space, including supernovae and supermassive black holes.
On Tuesday, he won $250,000 in the Regeneron Science Talent Search for his efforts.
Every year the competition casts a net across the nation for high schoolers doing the type of research you might expect from graduate students. This year Paz snagged first place out of nearly 2,500 entrants.
“Surprised isn’t a strong enough word,” Paz said shortly after the award ceremony. “I didn’t even give a thought to what I’d say to people if I’d won.”
The objects in Paz’s catalog aren’t just plain old stars or planets. They’re all variable objects, meaning they change dramatically, violently, and often unpredictably. A black hole, for example, can emit powerful jets that vary in brightness depending on how much material it’s gobbling up or how fast it’s spinning.
That makes these objects a wealth of information about some of the universe’s most befuddling mysteries. For example, they can be used to measure how quickly the universe is expanding from the Big Bang — a puzzle scientists are still trying to solve, which could rewrite physics.
Most of the objects Paz discovered are “candidates,” meaning further study is required for scientists to confirm what Paz’s analysis suggests they are.
Luckily, astronomers are already digging into his catalog.
Building an AI to scan the sky
Paz needed his machine-learning algorithm to comb through nearly 200 terabytes of data from a 10-year infrared survey of the entire sky by NASA’s NEOWISE space telescope.
Looking in the infrared — wavelengths invisible to the human eye — the NEOWISE mission searched for asteroids and comets near Eart. Infrared wavelengths, however, can also reveal objects deep in space that are shrouded in interstellar dust.
Even though NEOWISE wasn’t designed to look for such objects, Paz thought he could tease them out of the data with his AI algorithm.
“Prior to Matteo’s work, no one had tried to use the entire (200-billion-row) table to identify and classify all of the significant variability that was there,” Davy Kirkpatrick, who was Paz’s mentor at Caltech, told BI in an email.
Other surveys had tried to comb through NEOWISE data for specific types of variable objects, he added.
At the end of the summer program, “we were so impressed with his results that we hired him part-time at Caltech to finish the catalog,” Kirkpatrick said.
Paz said a lot of that work was him “in a dark room, eye bags heavy, looking at my computer, trying to solve a bug.” Sometimes he worked out math problems on a whiteboard at Caltech. He also consulted a variety of astrophysicists and astronomers.
Once the algorithm was ready though, it blew him away.
Making 1.5 million new discoveries
In order to analyze all 200 terabytes of data, Paz divided up the data into 13,000 equal parts.
The algorithm analyzed miniscule changes in infrared radiation to identify variable space objects and sort them into different classes, such as black holes or double-star systems. In some constellations, the algorithm was discovering more objects than anticipated.
“That was where I first started to see a lot of promise in the project,” Paz said.
In the end, he surveyed over 450 million objects in the sky and identified 1.9 million that may be variable objects like black holes or supernovae. Of those, 1.5 million had never been cataloged before — they were new discoveries.
“It’s very beautiful. Not just that number — it’s a big number that obviously makes you proud — but when you visualize the data,” Paz said.
Here’s that visualization, plotting all the candidate objects he discovered:
“You can see the Milky Way, you can see satellite galaxies, you can see Andromeda, you can see star-forming regions,” Paz said. “Even though it’s a very one-dimensional view of the universe, just plotting a point at every discovery we’ve made, we can really see the intricacies and the glory of the night sky.”
Now an infrared research group at Caltech is already using his catalog, called VarWISE, to study dual-star systems in the distant universe. They’ve already found dozens of star systems in VarWISE that weren’t previously detected, Kirkpatrick said. He added that the research helps them calculate the mass of distant alien planets.
Paz is submitting the catalog for publication in the Astrophysical Journal later this year. The catalog has not yet gone through the peer-review process, but the algorithm itself was peer-reviewed and published in the Astronomical Journal in November.
“The variable candidates that he’s uncovered will be widely studied and illustrate the enduring value of astronomical surveys,” Amy Mainzer, a scientist who led the NEOWISE mission, told BI in an email.
“It’s clear that he is simply a unique talent — smart, hardworking, and with a crazy ability to assimilate newfound knowledge into new ideas for studying the universe,” Kirkpatrick said.
From LA fires to the Big Bang
As for Paz and his $250,000, the next frontier is college. He said he’s been accepted at Stanford University, and is keeping his mind open about potential career paths.
Just weeks before flying to Washington, DC for the awards ceremony, Paz woke up in his Pasadena home to see flames outside the window. The Eaton fire traveled so quickly that he had received no official warning. After evacuations and several days of fire, his family’s home was spared.
“It really gives you a new perspective,” he said. “I have a new appreciation for the problems that I have the privilege not to worry about.”
Now he’s pondering the possibility of putting an infrared telescope into Earth orbit — this time to monitor Earth itself for emerging fires.
More immediately, though, Paz wants to use his NEOWISE findings to study the elusive expansion rate of the universe, starting from the Big Bang, and hopefully help scientists solve the biggest mystery in cosmology.
“It will either contribute to the resolution of a very contentious topic in current research, or it’s going to reveal something truly foundational about the origins of the universe,” Paz said.
