Burçin Mutlu-Pakdil Wins Schmidt Sciences Award to Hunt Dwarf Galaxies With AI

The astrophysicist is one of five inaugural FirstLight awardees given first access to a new generation of telescopes.

The next generation of telescopes will capture more of the night sky than any astronomer could examine alone. And hidden in the massive amount of data will be galaxies no one has seen before.

Burçin Mutlu-Pakdil, an assistant professor of physics and astronomy, is setting out to find them. As one of five early-career scientists in the first class of the Schmidt Sciences FirstLight Awards, she will receive up to $500,000 a year for up to five years—as much as $2.5 million total—to lead a team hunting for the smallest and faintest galaxies in the universe.

Left to right: Vivek Venkatraman Krishnan, Max Planck Institute for Radio Astronomy; Burçin Mutlu-Pakdil, Dartmouth College; Adina Feinstein, Michigan State University; Kimberly Ward-Duong, Smith College; and Griffin Hosseinzadeh, UC San Diego.

The inaugural FirstLight awardees. Left to right: Vivek Venkatraman Krishnan, Max Planck Institute for Radio Astronomy; Burçin Mutlu-Pakdil, Dartmouth College; Adina Feinstein, Michigan State University; Kimberly Ward-Duong, Smith College; and Griffin Hosseinzadeh, UC San Diego. (Photo courtesy of Schmidt Sciences)

Schmidt Sciences created the FirstLight Awards to put its new telescopes in the hands of early-career researchers. The five winners were chosen from more than 100 applicants through three rounds of outside review. FirstLight researchers will make their findings, software, and related work free and publicly available. The five teams will gather annually to trade methods, with awardees also leading workshops and public talks.

The four new telescopes—Argus, Lazuli, LFAST, and DSA—are known collectively as the Eric and Wendy Schmidt Observatory System. The instruments combine optical, infrared, and radio wavelengths to give them an observational range from nearby exoplanets to the large-scale structure of the universe.

“What makes this project particularly exciting is the opportunity to combine two complementary ways of searching for galaxies: optical observations, which reveal their stars, and radio observations, which trace the hydrogen gas that fuels star formation,” Mutlu-Pakdil says. “Some galaxies may be easier to identify through their stars, while others may first reveal themselves through their gas.”

The name of Mutlu-Pakdil’s project fits its scope: the Schmidt End-to-End Dwarf Discovery pipeline, or SEEDD.

“Our ultimate goal is not simply to discover more galaxies, but to build a much more complete picture of the faintest galaxy population,” she says. “How small can a galaxy be? Why do some tiny galaxies continue forming stars while others stop?”

‘Nearly Invisible Needles in an Enormous Haystack’

“Modern telescopes are producing astronomical images at a scale that is becoming impossible for humans to examine individually,” Mutlu-Pakdil says. “Somewhere within these enormous datasets are galaxies that we have never seen before.” Finding them, in her words, is “like searching for a handful of nearly invisible needles in an enormous haystack.”

Her team will build AI tools that “systematically search these images, recognize the subtle signatures of extremely faint galaxies, and identify promising candidates for further investigation,” she says. The software decides where the astronomers should look.

With earlier National Science Foundation funding, Mutlu-Pakdil built a first version of such an algorithm and used it to flag more than 100 possible dwarf galaxies, three of them now confirmed. One was named Kamino, after a planet in Star Wars, by Guinevere Herron, a Guarini PhD student in her group.

“This project will be a collaborative effort bringing together expertise in optical and radio astronomy, galaxy formation, and artificial intelligence,” Mutlu-Pakdil says. Her team, including postdoctoral researchers, graduate students, and undergraduates, will work closely with project co-leads David Sand, a professor at the University of Arizona, and Kristine Spekkens, a professor at Queen’s University, whose “complementary expertise will be essential to combining optical and radio searches for the faintest galaxies,” she says.

Mutlu-Pakdil is the second Dartmouth researcher in the past year to receive support from Schmidt Sciences for AI-driven work. In 2025, Jesse Casana, the W.J. Bryant 1925 Professor of Archaeology, won a grant through the organization’s Humanities and AI Virtual Institute to build open tools that find unknown archaeological sites in satellite and aerial imagery.

Ultimately, Mutlu-Pakdil hopes to discover clues about how the universe works. “What can these extreme systems tell us about dark matter and how galaxies form? These are the fundamental questions we hope to answer,” she says.

Written by

Arts and Sciences