🇺🇸 Matteo Paz

Matteo Paz grew up in Pasadena, California — a city whose skyline is dominated, intellectually, by the California Institute of Technology and NASA's Jet Propulsion Laboratory. Proximity became destiny. In 2022, while still an underclassman in high school, Paz enrolled in Caltech's Planet Finder Academy, an outreach program designed to give school students a taste of real astronomy.

A taste was not enough. Paz signed up for a six-week summer program at Caltech that pairs students with campus mentors, where he began working with astronomer Davy Kirkpatrick at IPAC, Caltech's astronomical data center. The assignment that would change his life involved NEOWISE — the repurposed Wide-field Infrared Survey Explorer spacecraft that had spent more than a decade scanning the entire sky in infrared light, primarily hunting near-Earth asteroids.

NEOWISE's archive was both a treasure and a problem. The mission had produced roughly 200 billion individual data entries covering nearly two decades of sky observations. Buried in that haystack were variable objects — stars that pulse, binary systems that eclipse each other, distant black-hole-powered galaxies that flare — whose brightness changes over time. Identifying them by hand, one light curve at a time, was simply impossible at that scale.

Paz's answer was to stop looking by hand. He developed a machine-learning algorithm — an AI model trained to recognize the subtle infrared signatures of variability in the raw data — capable of processing the entire NEOWISE archive and picking out sources whose brightness genuinely changed, as opposed to instrumental noise pretending to.

The results were staggering. His algorithm flagged approximately 1.5 million previously unknown potential celestial objects — candidate variable stars, black holes, and other astrophysical sources that had been sitting in NASA's own data, unseen, for years. In astronomy, where a single new discovery can justify a career, a catalog of 1.5 million candidates produced by a teenager was a genuine event.

The work was not a black box stunt. Paz and his mentor Kirkpatrick prepared the results for the professional community, with plans to publish the complete catalog of variable objects from the NEOWISE data so that astronomers worldwide could follow up on the discoveries. The project demonstrated a principle with implications far beyond one telescope: enormous scientific value lies dormant in archived data, waiting for better tools to release it.

In March 2025, Paz entered the Regeneron Science Talent Search, the competition (formerly sponsored by Westinghouse and Intel) that has been identifying America's most gifted young scientists since 1942 — its alumni include numerous Nobel laureates. Against 40 finalists whose projects ranged from treating rare muscle diseases to solving long-standing mathematics problems, Paz took first place and the $250,000 top prize.

Coverage of his win, from the Smithsonian to Caltech's own publications, emphasized the same point: this was not a student imitating research, but a student doing research at professional scale. The University of Arizona's Steward Observatory noted with some pride that his mentor Kirkpatrick was their PhD alumnus — a reminder that great prodigies are usually the product of talent meeting generous mentorship.

Paz has been open about the origins of his mathematical confidence, crediting an inspiring high-school calculus teacher for accelerating his love of the subject, and Caltech's programs for handing him a real problem instead of a toy one. His trajectory — from outreach-program student to author of a million-object catalog — took barely three years.

The technical challenge he solved deserves spelling out. NEOWISE was designed to hunt asteroids, not to catalog variable stars; its observations of any given patch of sky were episodic, noisy, and never intended for the kind of time-domain astronomy Paz attempted. Extracting genuine variability from that data meant teaching a model to distinguish a star's true flicker from instrumental artifacts, cosmic-ray hits, and the thousand ways a detector can lie. Professional surveys employ teams and years for such work. Paz compressed it into a high-school research project — and produced a candidate list large enough to occupy follow-up astronomers for a decade.

The prize money, substantial as it is, may prove the least valuable part of his win. The Regeneron Science Talent Search functions as American science's most reliable early-warning system: its top winners historically flow into the country's elite research institutions and laboratories, and Paz — already embedded at Caltech's IPAC before finishing high school — enters that pipeline with a professional-scale result already behind him. Astronomers quoted about the project made the point plainly: the catalog is not a student exercise; it is data the field will actually use.

What makes Paz's story resonate beyond astronomy is its method. He did not build a telescope; he built an instrument for reading what telescopes already saw. In an era when science generates data faster than humans can inspect it, the seventeen-year-old from Pasadena demonstrated that the next great discoveries may belong to whoever writes the best algorithms — and that there is no minimum age for doing so.