Embedded Files
Full-duplex wireless communication — the simultaneous transmission and reception of signals on the same frequency channel — is an emerging technology with the potential to dramatically improve the spectral efficiency of next-generation wireless networks. The central obstacle to its practical realization is self-interference: the signal a radio transmits is many orders of magnitude stronger than the signal it is trying to receive, and must be suppressed across wide bandwidths in real time. As part of the FlexICoN project at Columbia University's WiMNet Lab, this work develops a complete, real-time adaptive full-duplex radio system — integrating a state-of-the-art programmable IC analog canceller with a software-defined control and signal processing stack built in GNU Radio — and demonstrates its operation in experimental settings.
Whole Slide Images (WSIs) in pathology are enormous, rich in detail, and notoriously difficult to annotate at scale. While deep learning methods have shown promise in handling this complexity, they function as black boxes — lacking the interpretability necessary in the medical field. Developed as part of my M.Eng. thesis, this project introduces an unsupervised segmentation method for histopathological WSIs based on sparse dictionary learning. By clustering learned dictionary atoms based on shared activation patterns and reconstructing image patches through these sub-dictionaries, the algorithm assigns semantic labels with transparency. The method is demonstrated by segmenting glomerular structures extracted from mouse kidney tissue WSIs.
EpiWear is a compact, wrist-worn epinephrine injector designed to be truly wearable — minimizing size without compromising function. Developed at The Cooper Union in response to the life-threatening risks of anaphylaxis without immediate access to medication, EpiWear reimagines the traditional auto-injector in a sleek, watch-like form factor that fits seamlessly into everyday routines, addressing the critical gap in accessibility and compliance that existing devices leave behind. The invention has been recognized in multiple medical innovation competitions for its technical feasibility, safety-focused design, and potential for real-world impact.
Page updated
Google Sites
Report abuse