The dataset had been sitting there for years.
NASA's NEOWISE telescope spent a decade scanning the full sky in infrared, collecting nearly 200 billion rows of measurements. The data captured countless celestial objects — flickering quasars, pulsing binary stars, fading supernovae — but most of them had never been found. The archive was simply too vast, too noisy, and too complex for standard tools to pick through.
Then a high school student from Pasadena decided to take a look.
Matteo Paz, a student at Pasadena High School in California, joined the Planet Finder Academy in the summer of 2022 — a programme that places students inside real-world astronomy challenges under the mentorship of working scientists. Paired with Caltech scientist Davy Kirkpatrick at NASA's Infrared Processing and Analysis Center (IPAC), Paz was given access to the NEOWISE archive and tasked with studying a small subset of objects manually.
Instead, he built something bigger.
Drawing on his background in theoretical mathematics, coding, and time series analysis, Paz designed an automated algorithm to process the entire archive. In six weeks, he created a machine-learning pipeline capable of detecting faint, variable light sources — objects whose brightness changes too subtly or unpredictably for humans or standard software to detect.
"The model started showing promise almost right away," Kirkpatrick told Phys.org. "As Paz fine-tuned it, the results only got more exciting."
The system used Fourier transforms and wavelet analysis — mathematical tools for studying time-based signals — to reveal faint variations in the infrared spectrum that NEOWISE's standard processing had missed. Some objects changed so slowly or briefly that they'd been invisible in the data for over a decade.
When Paz ran his pipeline across the full archive, it identified 1.5 million previously unknown objects.
One-point-five million.
His findings were published in The Astronomical Journal — a peer-reviewed scientific publication — making Paz one of the few high school students ever to publish a paper in a major astrophysics journal.
The discovery highlights something profound about the current moment in science. The data is often already there. The universe has been observed, catalogued, archived. What's missing is the intelligence to interpret it — and increasingly, that intelligence can be built by anyone with the right curiosity and tools. A high schooler. A summer programme. Six weeks.
1.5 million objects, discovered.
The universe is bigger than we knew. And apparently, so is the talent pool. 🌌