High School Student's AI Model Uncovers 1.5 Million Hidden Cosmic Objects in NASA Data

Matteo Paz, an 18-year-old from Pasadena, California, has transformed a massive archive of noisy telescope data into one of the most significant astronomical catalogs of the decade. Using a custom-built artificial intelligence algorithm, the high school senior uncovered 1.5 million previously unknown cosmic objects, ranging from eclipsing binary stars to distant quasars.[1][2]

The breakthrough centers on data collected by NASA's Near-Earth Object Wide-field Infrared Survey Explorer (NEOWISE). Launched in 2009, the space telescope spent over a decade scanning the sky in infrared light. While its primary mission was to track near-Earth asteroids, NEOWISE inadvertently captured the faint, fluctuating heat signatures of countless distant phenomena.[4][6]

Over its operational lifespan, NEOWISE generated a staggering dataset containing nearly 200 billion rows of measurements. The sheer volume and chaotic nature of the data made it virtually impossible for human researchers to manually identify "variable objects"—celestial bodies whose brightness changes unpredictably over time. For years, this cosmic treasure trove sat largely untouched in digital archives.[1][4]

Enter VARnet, a machine-learning pipeline designed by Paz during a mentorship program at the California Institute of Technology's Infrared Processing and Analysis Center (IPAC). Drawing on advanced mathematics, including Fourier transforms and wavelet analysis, Paz trained his model to detect microscopic variations in the infrared spectrum that standard software missed.[2][4]

The algorithm first grouped scattered light observations belonging to the same star, filtering out the surrounding cosmic noise. Once the data was cleaned, the AI analyzed the time-based signals to reconstruct full light curves, successfully flagging 1.5 million objects that flickered, pulsed, or faded.[1][5]

In a rare achievement for a high school student, Paz published his methodology and findings as the sole author in the peer-reviewed Astronomical Journal. The resulting catalog, dubbed VarWISE, was quickly embraced by the professional astronomical community, proving that the AI's detections were highly accurate and scientifically valuable.[2][4]

The discovery earned Paz the $250,000 first-place prize in the 2025 Regeneron Science Talent Search, the nation's most prestigious STEM competition for high school seniors. His work also caught the attention of major space institutions, earning formal praise from NASA leadership for unlocking the hidden value in their retired telescope's archives.[1][2][4]

The impact of the AI model is already being felt at major observational facilities. In early 2026, the Vera C. Rubin Observatory in Chile integrated Paz's catalog into its real-time alert system. The observatory is now using the data to flag exploding stars and active black holes, issuing hundreds of thousands of alerts to astronomers worldwide.[1][3]

Beyond the astronomical discoveries, the project highlights a massive shift in modern scientific research. Tools that previously required institutional supercomputers and entire university departments are now accessible to young students with a laptop, open-source data, and a fresh perspective.[1][6]

Paz and his mentors believe the underlying architecture of VARnet has applications far beyond the stars. Because the algorithm is fundamentally designed to analyze time-series data, it could eventually be adapted to track subtle environmental changes, monitor financial markets, or decode complex neurological signals.[2][4]