Dr. Binxu Wang
Incoming Group Leader, Janelia Research Campus (Dec 2026)
Research Fellow, Kempner Institute at Harvard University
PhD, Washington University in St. Louis
Dr. Binxu Wang first encountered neuroscience through science fiction. In The Dark Forest, the second novel in Liu Cixin’s The Three-Body Problem trilogy, a fictional neuroscientist is recruited to save humanity from an alien invasion. His radical vision is that unwavering beliefs can be implanted into the human mind through a sufficiently deep understanding of the brain. Beneath the science fiction lies a provocative premise—that the brain’s internal computations could one day be read, modeled, and even rewritten. This idea, however fanciful it seemed at the time, utterly captivated Binxu. That idea would gradually guide her towards a career at the intersection of neuroscience and AI—another endeavor that was confined to science fiction not so long ago. As a Research Fellow at the Kempner Institute at Harvard University and a soon-to-be Group Leader at Janelia Research Campus, Binxu’s research is revealing how representations emerge both in artificial neural networks and in living brains.
For a time, Binxu set aside her dream of understanding—and ultimately simulating—the brain. Inspired in part by The Three-Body Problem, she became convinced that physics offered the deepest route to reality, and chose to pursue an undergraduate degree in physics. Still, her dormant interest in neuroscience would periodically resurface. While preparing a final presentation for a computer science seminar at Peking University, she became captivated by artificial neural networks and chose to present a landmark paper from the 80’s on backpropagation. This was 2013, just before backpropagation would be rediscovered and celebrated as the engine behind the coming deep learning revolution. But Binxu was ahead of the curve, and the significance of that work was not yet appreciated by her classmates or her professor. Disheartened, Binxu decided to focus fully on physics and math.
But a few years into her undergraduate degree, Binxu had an existential crisis about her chosen path. Reading Thomas Kuhn’s The Structure of Scientific Revolutions challenged her deeply held conviction that equations were somehow more fundamental than the world they described. She began to see how lonely this viewpoint was, as other scientists like biologists and even mathematicians held the more nuanced view that equations were useful models of the world, but not the world itself. Much as she loved the elegance of describing the world through mathematics, physics no longer seemed uniquely privileged among the sciences, and she wanted whatever she studied to relate more directly to her own life and the broader human experience. Binxu thus decided to change course and return to the fascination that had first inspired her: understanding the brain.
After a short research experience in psychology, Binxu decided that she still didn’t want to completely abandon physics, but instead to harness it as a useful tool for understanding the brain. She joined the lab of Dr. Louis Tao, who was one of relatively few theoretical and computational neuroscientists in China at the time. He also happened to have a physics and math background, much like hers. He acknowledged that trying to fit the pristine equations of math and physics to the messy realities of biology could often yield disappointment. Instead, he encouraged her to embrace biological problems and data rather than force existing mathematics upon them. The most enduring mathematical ideas, he argued, often emerge from attempts to solve real-world problems. Biology might inspire new mathematics, revealing insights that would otherwise remain hidden. This was a revelation for Binxu that changed how she thought about science and would continue to guide her for the rest of her career.
Binxu knew she wanted to go to graduate school and focus on computational neuroscience. Following some summer research experiences with Dr. Alex Reyes at New York University and a particularly influential conversation with Dr. Eero Simoncelli while interviewing for NYU’s graduate program, she became convinced NYU was the perfect fit. Unfortunately, she was waitlisted, and then rejected with less than 24 hours left to choose among her remaining offers. She ultimately made the decision to attend Washington University in St. Louis based on little more than a gut instinct. Although it was the choice furthest outside her comfort zone, it turned out to be a good one. While WashU was home to more experimental than computational neuroscientists, a brand-new faculty member, Carlos Ponce, had just arrived and turned out to be an ideal fit for her interests. His proposal of using deep neural networks to generate novel images to optimally stimulate the brain sounded like the sort of science fiction-esque neuroscience that had always captivated her. The only catch: his lab conducted technically difficult electrophysiology experiments with macaque monkeys, and Binxu had never so much as touched a single animal in her research training. Once again, Binxu embraced some risk and took on the challenge.
For her PhD in the Ponce lab (which eventually moved from WashU to Harvard), Binxu studied how the primate visual system represents complex visual stimuli. She used deep neural networks to generate visual stimuli that were presented to a monkey and drove electrophysiologically recorded neuronal activity. That activity was then used as feedback to guide the network to iteratively generate new images, ultimately creating the stimulus that maximally excited the recorded neurons. Through this work, Binxu found that neurons' stimulus tuning can be visualized as a landscape in a high-dimensional image space, where each image is a point and the neuron’s firing rate defines the elevation. The neuron’s preferred stimulus is the summit of this landscape—a mountain peak that the optimization algorithm gradually climbs. Across the ventral visual pathway, these peaks become progressively narrower yet higher-dimensional, reflecting neurons that respond more selectively while integrating increasingly rich combinations of visual features. For Binxu, this offered a geometric view of neural representations, reconnecting the mathematical beauty she had always admired with the messy reality of the brain.
While making these exciting discoveries, Binxu developed considerable expertise with generative models. This turned out to be timely, as large language models (LLMs) and text-to-image diffusion models—two of the most visible forms of modern generative AI—soon burst onto the scene and into the public consciousness. Binxu initially sought a research topic that felt closer to human experience, and she found one with exceptional relevance—not only for how individuals see and perceive the world around them, but also for society at large. Her expertise was increasingly in demand, so when she moved from the relatively affordable St. Louis to the far more expensive Boston, financial realities led her to seriously consider a career in industry. She applied to many Neuro-AI jobs at Google, Meta, and other major tech companies, but visa complications ultimately prevented any of these opportunities from materializing. Though initially discouraged, a new opportunity came to light as the Kempner Institute for the Study of Natural and Artificial Intelligence was founded at Harvard in 2023. The Kempner Institute advertised a handful of unique Research Fellow positions for recent PhD graduates. Research Fellows have more independence than postdocs but are on shorter-term (3 year) contracts relative to standard tenure-track positions. Binxu was a perfect fit and was accepted into the Kempner’s inaugural class of Research Fellows. Obtaining this job strengthened Binxu’s confidence, gave her the space to develop an independent research program, and solidified her goal of pursuing a long-term academic career.
Today, Binxu is preparing to move to a longer-term faculty position at HHMI’s Janelia Research Campus. Her research group will continue studying how generative models work and how they can be used to disentangle the complexity of the brain. Binxu’s work perhaps bears closer resemblance to that of the fictional neuroscientist in The Three-Body Problem than she might ever have imagined. Yet her work is deeply rooted in biology, and she is an emerging leader in the quest to apply mathematical theory to understanding the human experience—the same quest that drew her to neuroscience in the first place.
Find out more about Binxu and her lab’s research here.
Listen to Melissa’s full interview with Binxu on April 27, 2026 below!
