I'm an undergraduate biomedical engineer (B.S., expected May 2028; GPA 3.83, Dean's List) working at the intersection of flexible electronics fabrication, deep learning, and dexterous robotics.
I like building the full signal chain of an instrument — fabricating the electrode, writing the firmware that drives it, and training the model that makes sense of what comes back. I am currently working as an undergraduate reseearch at the Columbia Laboratory for Unconventional Electronics (CLUE) under Dr. Ioannis (John) Kymissis.
Electrodes, flexible arrays, embedded firmware, and the benchtop rigs that prove they work.
Imaging pipelines, 3D deep learning, multimodal fusion, and classical ML that turns signal into insight.
A wearable system using flexible LED–photodiode arrays to non-invasively screen for breast cancer by reconstructing optical absorption through tissue. Wrote STM32 firmware (STM32CubeIDE) for individual LED addressing, and optimized illumination/detection patterns to maximize source–detector coverage and SNR.
A four-electrode sensor for ventriculoperitoneal shunts: an upstream pair pulses the conductive CSF and the downstream pair times the ion cloud's transit, mapping to flow rate and revealing obstruction. Built the benchtop loop (NE-1000 syringe pump, artificial CSF, Keysight WaveGen + capture), characterized Pt–Ir electrodes via EIS, and designed a translational version with cleanroom metal deposition + Parylene-C.
Conductive ink drawn directly on skin to capture EEG, ECG, and skin-impedance signals for stress detection and wearable diagnostics. Developing flexible DoS transistors, heaters, and pressure sensors, and mentoring incoming researchers in lab protocols.
Led 7 undergraduates applying AI/ML to brainwave data from commercial EEG headsets. Streamlined acquisition with an IRB-approved SOP over Lab Streaming Layer, and classified grasping motions with a Gaussian Mixture Model + K-Nearest Neighbors.
SFCN (3D CNN) and a Vision Transformer with a 3D-CNN tokenizer stem, trained on 712 ADNI subjects with paired T1 + FLAIR MRI. Full preprocessing (N4 bias correction, MNI152 registration) and transfer learning from a UK Biobank checkpoint.
Integrating SpikeATac piezoelectric tactile sensors onto the fingertips of CRAFT, an open-source tendon-driven hand with hybrid hard–soft compliance, then designing a multimodal architecture that fuses touch + vision to train manipulation policies.
Designed and tested a levitating sole using neodymium magnets to relieve plantar fasciitis — achieving 84.57 lbs of repulsion, an improvement over traditional EVA foam. Applied K-Means clustering to plantar-pressure and gait data to tune magnetic force across the sole for optimal redistribution.
Yu, Y. (2024). Levitating Magnetic Insoles: A Novel Approach to Alleviating Plantar Fasciitis Through the Reduction and Redistribution of Plantar Pressures. Journal of Innovations in Medical Research, 3(3), 14–35.