Motivation

Engineering the Mind

How do patterns of neural firing give rise to the ability to distinguish right from wrong, develop emotions, store memories, and experience consciousness? These questions, which once felt philosophical, have evolved into research ideas I have been trying to pursue for the past three years. Yet the deeper I venture into neuroscience, the more it dawns on me that these questions might not be answered in my lifetime. Our most advanced neural imaging tools only scratch the surface, offering fragmented glimpses of an immensely complex system. I have worked alongside highly skilled and motivated researchers and can testify that the field is not short of intelligence and effort. Instead, the field is in dire need of a revolutionary unifying framework or a transformative technology capable of capturing brain activity at its full scale and speed. For instance, what would it mean if we could noninvasively visualize every neuron’s activity in real time? To many, that idea seems as impossible today as landing on the moon in the 1600s, as Professor Ed Boyden once remarked. But for me, it is the kind of audacious question that fuels my curiosity.

My research journey began in my freshman year after joining a terpene biosynthesis lab. There, I grappled with expressing and purifying membrane-bound UbiA terpene synthase enzymes to uncover novel terpenoids. Using a geranylgeranyl diphosphate (GGPP)-overproducing E. coli system, I screened 32 candidate enzymes and identified five that produced structurally diverse diterpenes. This work culminated in three publications in Nature Communications , ScienceDirect , and ACS Catalysis . Presenting this research at the 2023 American Society for Pharmacognosy conference and an international symposium further strengthened my scientific communication skills. Though outside my primary field of interest, this experience taught me the rigor and resilience required for experimental science. Witnessing a simple hypothesis evolve into peer-reviewed articles ignited an insatiable desire to dive deeper into research.

Despite success in biochemistry, my curiosity gravitated toward the brain. In my sophomore year, I joined a deep brain stimulation (DBS) clinical research lab at UF. Transitioning from biosynthesis to computational neuroscience demanded rapid skill acquisition, which I undertook enthusiastically under great mentorship. My first task was labeling video data of Tourette patients into tic and non-tic segments–a process that quickly became tedious, taking over two hours to annotate a 15-minute clip. To improve efficiency, I pitched the idea of building an automated tic detector. My PhD mentor supported it but cautioned against challenges such as limited training data and high tic variability across subjects. To address these, I proposed integrating concurrent electromyography (EMG) signals into the training pipeline to strengthen model performance. Thereafter, I performed a comparative tic detection accuracy analysis on the two modalities and submitted a first author manuscript for publication. I was inspired by how DBS, despite incomplete theoretical understanding, can meaningfully improve patients’ lives. I will never forget one participant who, at the study’s start, could barely suppress tics for 30 seconds but, two years later, had improved enough to drive independently for the first time. That experience revealed to me that while understanding the brain is beautiful, restoring function is deeply human. Yet it also left me wondering: how much more could we achieve if we truly understood how DBS works connectomically? That question led me toward fundamental neuroscience research.

Therefore in the summer of 2024, I joined Professor Mala Murthy’s lab at the Princeton Neuroscience Institute to study visual processing in Drosophila . I performed computational analysis on FlyWire electron microscopy connectome data. Specifically, I probed the structure-function relationships in Lobula Columnar (LC)11 neurons, which detect small-object motion. Results from my analysis challenged the existing hypothesis of center-surround inhibition–where neurons respond strongly to small objects but are inhibited by larger stimuli extending beyond receptive fields. Surprisingly, the data revealed an overwhelming predominance of inhibitory synapses, contradicting previous assumptions and opening exciting avenues for new theories. Presenting these findings at the 2024 Society for Neuroscience conference reinforced my appreciation for integrating computational theories with experimental validation. Yet I realized that connectomic data, while powerful, is static. A connectome of a dead fly cannot fully explain the dynamics of a living one. That realization drove me toward technologies capable of capturing the live computation of the brain.

In light of this observation, I interned at MIT under Professor Laura Lewis through the MIT Summer Research Program 2025. Here I was tasked with mapping key brain metabolites fluctuations across sleep stages among healthy controls and major depression disorder subjects. I segmented EEG recordings into sleep stages and used them to segment concurrently recorded HERCULES-edited MRS data to optimize signal-to-noise ratio. Shadowing MRI and EEG data collection exposed me to the complexity of human neuroscience: every signal–noisy, variable, and often ambiguous–carries both promise and limitation. EEG provides temporal precision but poor spatial resolution. MRS gives chemical insight but averages across vast tissue volumes. Confronted with these constraints, I initially wondered whether better computational models alone could resolve the gaps. But I realized that when we sample only a tiny fraction of a highly interconnected brain, no algorithm can fully recover what was never measured. We have become experts at capturing fragments, yet our theories struggle to unify them. This experience solidified my belief that transformative progress demands advances in both measurement technology and modeling–and especially tools capable of capturing whole-brain dynamics directly.