Edited from the original story at the College of Science.
Berton Earnshaw, recently appointed as a research professor at University of Utah, is no stranger to the flagship university of the Beehive State where he has been an adjunct professor in mathematics since 2018. He also earned his PhD in mathematics at the U in 2007 and served as senior fellow for the College of Science since July 2024. At that time Peter Trapa, dean of the college, stated that Earnshaw was “at the top of his game in machine learning as it relates to drug development.”
Earnshaw will be leaving his position at Recursion, where he was a founding fellow, and setting up shop as a research professor with joint appointments in mathematics and biology. This transition back to academia promises considerable benefits for graduate students, post-doctoral researchers, and fellow faculty at both the main campus and U of U Health. Among those with the most to gain are U students and researchers in fields related to “techbio,” where AI and other advanced technologies are revolutionizing life sciences industries and improving lives in real time. Beyond the lab, Earnshaw’s deep experience in commercialization stands to strengthen the broader life sciences and tech communities at the U. His involvement in programs like the Science Research Initiative (SRI) and the emerging bioinformatics major will allow students to build their skills in this growing industry, and his own research endeavors will only add to these opportunities.
Spiraling upwards
Earnshaw’s career journey has been more than circuitous; it’s a career double helix spiraling upwards in which, invoking the philosopher Aristotle, the whole is greater than the sum of its parts. One part was his arrival at the U for a PhD, which he completed in record time: three years.
“I was very naïve,” he recalled. “I was thinking…I want to do math and model consciousness.” His advisor Paul Bressof, now at Imperial College London, was doing neuroscience and Earnshaw thought that “this must be the way to get at this, and Paul also has a very philosophical bent, and we got along, and we did a lot of modeling, but it was all biophysics.”
In the end, there was no modeling of consciousness, said Earnshaw, though his models of protein-trafficking focused on synapses during episodes of learning and memory formation.
Following a post-doc at Michigan, Earnshaw found himself at a string of startups before arriving at Recursion as director of data science research in 2017. Previous stops included CTO of Perfect Pitch (now Boomsourcing), director of data science and operations at Red Brain Labs (acquired by Savvysherpa) and principal and senior scientist at Savvysherpa (acquired by UnitedHealth Group).
Earnshaw has also served as a member of the Utah State Auditor’s Commission on protecting privacy and preventing discrimination.
Building more sensitive ‘maps’
For the past several years at Recursion, a Salt Lake City-based drug discovery startup, Earnshaw has been pioneering a radical approach to pharmaceutical development—one where data, not intuition, determines which diseases to pursue. Using fluorescent and brightfield microscopy to capture images of cells across “basically the entire human genome,” his team has built “maps of biology” that fundamentally reshape drug discovery economics. While a researcher might observe that “here’s a healthy cell, here’s a diseased cell, and you can kind of see the difference,” Earnshaw explained that “that’s just a qualitative visual understanding.” The breakthrough lies in quantification: “using modern machine learning and AI, you can actually learn those representations. And they’re far more powerful in terms of being able to capture relevant information and allow you to build more sensitive maps of how, for example, genes and drugs are related to each other.”
This data-driven approach inverts traditional pharmaceutical research.
“We don’t decide up front, you know, hey, we already know a lot about this disease area, so that’s what we’re going to go after,” Earnshaw noted. “It’s the data that we generate that shows us where to go.”
The AI systems he’s led extract biological information from cellular morphology to reveal gene relationships, enabling Recursion to pursue both well-studied diseases and rare conditions pharmaceutical companies have traditionally ignored. “If you can change the economics of drug discovery, then you can go after these.”
By identifying promising drug candidates faster and more efficiently, Recursion aims to “speed it up, get to the ideas fast, prove them fast” with maximum signal per research dollar spent—making treatments viable even for patient populations once considered too small to justify investment.
Returning to foundational questions
Now, returning to the U as a research professor, Earnshaw will pursue two ambitious goals that merge his industrial experience with his foundational mathematical background: building comprehensive cellular models and reimagining AI itself.
“You really don’t understand a thing until you can actually simulate it,” Earnshaw reflected, articulating his first objective. With better microscopes and new measurement techniques generating unprecedented data about cellular activity, he aims to create predictive models of entire cells—moving beyond the “black box” approaches that dominated his industry work. By combining mechanistic modeling techniques from his Ph.D. days with modern machine learning, he hopes to achieve “not only better predictive models, but to do it in a way that gives us deeper understanding of what is actually happening in a cell.”
His second program tackles AI’s foundations directly. While large language models like ChatGPT have made “amazing” progress, Earnshaw believes they lack true world models—they’re “very good at mimicking in a conditional way” rather than reasoning from internal understanding. “The better we can understand that difference,” he argued, “the better we are going to be able to build AI systems” that serve humanity rather than threaten it.