I'm an enthusiastic researcher looking to apply my computational skills as an electrical engineer to have a biological impact. I seek to use machine learning, statistics and mathematical methods to answer questions in biology and neuroscience. My main areas of focus are computational biology for cancer research and neural processing in brain computer interfaces.
My research has been supported by the following institutions
I'm currently working as a research assistant in the Christina Leslie lab at Sloan Kettering Institute, where I'm designing a statistically inspired robust pipeline for analysis of flow cytometry data of renal cell carcinoma patients. I'm also simultaneously working in the Sudhin Shah lab at Weill Cornell Medicine, in collaboration with Prof. Nima Mesgarani at Columbia University, where we are using temporal response functions to analyse EEG data of patients suffering from acute cognitive impairment. Previously, I have also been a part of the Bionet Lab at Columbia University working with Prof. Aurel Lazar on graph-based models for integration of landmark and motion information in the Drosophila fruit fly.
I was awarded the Nikola Tesla Scholarship at Columbia University for an admission into the Master of Science (MS) program in Electrical Engineering. During the course of my undergraduate studies at IIT Kharagpur, I was awarded the Best Bachelor's Thesis in Electrical Engineering for my work in physical layer secrecy and also the Nilanjan Ganguly Memorial Endowment for academic excellence.
Beyond career ambitions, I have a great passion for music and the arts - I play the guitar, piano and violin. I'm also an avid reader and a movie/TV buff, believing in the philosophies presented by larger-than-life shows like Community & Doctor Who.
A. Sinha, P. Mohapatra, J. Lee and T. Q. S. Quek, "On the Secrecy Capacity of 2-user Gaussian Interference Channel with Independent Secret Keys," 2018 International Symposium on Information Theory and Its Applications (ISITA), Singapore, 2018, pp. 663-667. doi: 10.23919/ISITA.2018.8664253 [IEEE] [pdf]
We use graph-based louvain clustering and sparse gaussian mixture models to cluster high-dimensional flow cytometry data into biologically plausible populations with distinct phenotypes. This pipeline seeks to replace manual gating as the new norm for flow analysis, yielding more robust and reproducible results. Read More...
EEG recordings are taken from a dry electrode kit in response to a language listening task of an excerpt from Alice in Wonderland, by Lewis Caroll. The audio is played both forwards and reversed and various methods like spectral analysis, temporal response functions and semantic word embeddings are employed in an effort to probe higher cognitive function. Read More...
We use neural responses to speech from EEG and pre-learned spectro-temporal receptive fields (STRFs) as the sensing matrix in a sparse formulation to estimate the input heard speech spectrogram, using methods like projected subgradient, accelerated proximal gradient, augmented lagrangian and Frank-Wolfe. Read More...
A combination of angular velocity edge-detection inputs and broad receptive fields from the retina is used as neural inputs to a ring oscillator based circuit in the ellipsoid body to track visual stimulus in real time. The neural circuit seeks to replicate the model and results of this paper on the graph-based Neurokernel framework developed by the Bionet Lab. Read More...
In this Kaggle challenge, we attempt to use target protein images and reference images of nuclei, endoplasmic reticulum and microtubulues to do multi-class classification of protein localization labels, using three parallel VGG-19 inspired CNN networks. Read More...
