Through his investigations at Caltech, a local high school scholar uncovered 1.5 million hitherto undiscovered entities in space, expanded the possibilities of a NASA undertaking, and published a solo-authored, peer-reviewed article.

Matteo (Matthew) Paz’s Astrophysical Journal paper details a novel AI algorithm he created that led to these revelations and that can be utilized by other astronomers and astrophysicists for their investigations. 

In recognition of these accomplishments, Paz, a senior at Pasadena High School, received the $250,000 top award in the Regeneron Science Talent Search managed by the Society for Science.

Paz has had a desire to explore more about astronomy ever since his mother took him to public Stargazing Lectures at Caltech during his elementary years. In the summer of 2022, he visited the campus to study astronomy and its related computer science disciplines in the Caltech Planet Finder Academy, conducted by Professor of Astronomy Andrew Howard. In 2023, he enrolled in Caltech’s six-week Summer Research Connection, another initiative managed by the Center for Teaching, Learning, and Outreach that connects local high schoolers with mentors in campus laboratories.

Astronomer and IPAC senior scientist Davy Kirkpatrick acted as Paz’s guide. Kirkpatrick has mentored high school students for the past five summers, in addition to an undergraduate student, citizen researchers, and visiting graduate fellows.

“I’m incredibly fortunate to have met Davy,” Paz remarks. “I recall the first day I spoke to him, I mentioned that I was thinking about developing a paper from this, which is a much larger ambition than six weeks. He didn’t dissuade me. He said, ‘OK, so let’s discuss that.’ He facilitated an uninhibited learning journey. I believe that’s why I’ve advanced so much as a scientist.”

Kirkpatrick was raised in an agricultural community in Tennessee and achieved his aspiration of becoming an astronomer thanks to his ninth-grade chemistry and physics instructor, Marilyn Morrison. She informed him and his mother that he possessed potential and outlined the courses he should take to prepare for higher education.

“I aimed to offer that same form of mentorship to someone else, and ideally, many others,” Kirkpatrick expresses. “If I detect their potential, I want to ensure they are fulfilling it. I’ll do everything I can to assist them.”

Kirkpatrick additionally sought to extract further understanding from NEOWISE (Near-Earth Object Wide-field Infrared Survey Explorer), a now-decommissioned infrared telescope that had surveyed the entire sky in search of asteroids and other entities near Earth for over a decade. While the NASA telescope was busy observing asteroids, it also identified the varying heat of other more distant cosmic entities that flashed brightly, pulsed, or dimmed as they were obscured. Astronomers refer to these variable entities as hard-to-detect phenomena like quasars, supernovae, and binary stars eclipsing one another. However, data on these variable entities had yet to be utilized. If the NEOWISE team could pinpoint those entities and make them accessible to the astronomical community, the resultant catalog could provide insights into how these cosmic entities evolve over time.

“At that juncture, we were approaching 200 billion entries in the table of every single detection we had made over the span of more than a decade,” Kirkpatrick states. “So my ambition for the summer was to take a small segment of the sky and see if we could locate some variable stars. Then we could bring those to the attention of the astronomical community, saying, ‘Here’s some new stuff we uncovered manually; just think about the possibilities in the dataset.’”

Paz had no plans of manually sifting through the data. His academic preparation equipped him to bring a fresh perspective to the task. He had developed an interest in AI during an elective that combined coding, theoretical computer science, and formal mathematics.

Paz understood that AI learns most efficiently from extensive, orderly datasets like the one Kirkpatrick had provided him. Additionally, Paz possessed the advanced mathematical skills necessary to enjoy coding: He was already studying higher-level undergraduate mathematics in Pasadena Unified School District’s Math Academy, where students complete AP calculus BC by the eighth grade.

Thus, Paz embarked on creating a machine-learning method to examine the entire dataset and flag potential variable entities. Over the course of those six weeks, he initiated the drafting of the AI model, which began demonstrating promise. As he progressed, he conferred with Kirkpatrick to grasp the pertinent astronomy and astrophysics knowledge.

“Every meeting with Davy comprises 10 percent work and 90 percent just us chatting,” Paz shares. “It’s been incredibly enjoyable just to have someone to discuss science with in that manner.”

Kirkpatrick also linked Paz with Caltech astronomers Shoubaneh Hemmati, Daniel Masters, Ashish Mahabal, and Matthew Graham, who imparted their expertise in machine-learning methodologies for astronomy and in studying objects that vary over short and long timescales. Paz and Kirkpatrick discovered that the specific rhythm of NEOWISE’s observations meant that it would not systematically identify and categorize many objects that either flashed rapidly or changed gradually over an extended period.

As the summer came to a close, ample work remained. In 2024, Paz and Kirkpatrick collaborated once more, and this time, Paz served as a mentor for other high school students.

Currently, Paz has enhanced the AI model to process all of the raw data from NEOWISE’s observations and has scrutinized the results. Trained to notice subtle differences in the telescope’s infrared readings, the algorithms flagged and categorized 1.5 million potential new entities in the data. In 2025, Paz and Kirkpatrick intend to publish the complete catalog of objects that exhibited significant variations in brightness in the NEOWISE dataset.

“The model I implemented can be utilized for other time domain studies in astronomy, and possibly any other field that involves temporal formats,” Paz states. “I could perceive some relevance to (stock market) chart analysis, where the information similarly comes in a time series and periodic components are essential. You could also examine atmospheric effects such as pollution, where seasonal patterns and day-night cycles play substantial roles.”

Paz’s experience at the Regeneron Science Talent Search imparted valuable lessons to Kirkpatrick about mentorship. “When they announced Matteo as the champion of the science competition, that was the highest peak I’ve ever experienced,” he recalls. “I’ve received accolades in the past as well, and that’s thrilling too, but when you assist someone in reaching a portion of their potential and they are recognized for it, it’s a wonderful feeling.”

Kirkpatrick adds: “The degree to which we can engage with the local community rich in intelligent young minds, mentor them, and ensure they don’t forget their potential, the better off we all are.”

Currently, as he approaches graduation from high school, Paz is employed by Caltech. He works under Kirkpatrick at IPAC, which manages, processes, archives, and analyzes data from NEOWISE and several other NASA and NSF-supported space missions. This marks Paz’s first paid position.