Academic research and course projects.

End-to-end software pipeline for neural data analysis using Neuropixels probes

Visiting Research Scholar

Purdue University · West Lafayette, IN

• Accelerated the analysis of neural data from Neuropixels probes by 10× by building a comprehensive custom Python software package.
• Explored deep learning approaches to develop computational models describing visual perception in the mouse brain.

Probing memory encoding via theta oscillations in deep neural networks

Graduate Research Assistant (M.S. thesis)

Purdue University · West Lafayette, IN

• Applied principles from predictive coding theory in a deep neural network to replicate a wet-lab experimental finding.
• Implemented a deep learning model that elicits oscillations in the theta frequency band (4–8 Hz) and analyzed power changes for familiar vs. novel images.

Thesis

Studying the learning impairments in FragileX mice

Graduate Research Assistant

Purdue University · West Lafayette, IN

• Developed machine-learning classifier models that predict stimuli and mice genotypes from neural signals recorded with Neuropixels probes.

Thesis

Speech decoding using ECoG signals

Graduate Researcher

Purdue University · West Lafayette, IN

• Devised a preprocessing pipeline that computes time-aligned phonemes from recorded speech data, used to train a speech decoder.

Automated extraction of ECG signals from PDFs

Undergraduate Research Intern

BEES Lab · Indian Institute of Science, Bengaluru

• Developed an algorithm to automatically convert non-overlapping 12-lead ECG signals recorded in the form of PDFs to numerical data.

Automated epilepsy classification using EEG signals

Undergraduate Research Intern

BEES Lab · Indian Institute of Science, Bengaluru

• Developed a fully automated epilepsy classification system to classify epileptic patients into various subtypes of epilepsy using EEG signals.

Publication

Bladder cancer segmentation using deep learning

Undergraduate Researcher

NIT Karnataka · Surathkal, India

• Implemented a deep learning neural network to segment immunopositive and immunonegative tumour cells in the bladder and computed its Ki-67 index.

Code

Small signal parameter extraction of GaN and GaAs transistors

Summer Research Intern

CeNSE · Indian Institute of Science, Bengaluru

• Implemented a method to extract the parasitic capacitances, inductances and resistances from the S-parameters of GaN and GaAs transistors.
• From the extrinsic parameters, the intrinsic parameters of the MOSFET were further extracted.

Report

Vision transformers with learnable resizer networks

Purdue University

• Implemented and compared the performance of Vision Transformer models after integrating learnable resizer modules.
• Assessed the model on two datasets: Beans (0.97) and a subset of CIFAR-10 (0.95).

Code Report Video

Speech enhancement using CNN-GAN

National Institute of Technology Karnataka

• Evaluated a time-frequency mask based approach for speech enhancement through convolutional GANs (CNN-GAN).
• Compared speech quality metrics: mask (PESQ 2.34) vs. non-mask (PESQ 2.05) models.

Code Paper

Heart-chamber segmentation using deep learning

National Institute of Technology Karnataka

• Designed and implemented a two-chamber echocardiogram segmentation algorithm using deep learning, comparing against existing architectures on standard metrics.
• Demonstrated improved generalization over baseline architectures on held-out data.

Code Paper

Thirsty crow line-following robot

E-Yantra (2018), Indian Institute of Technology, Bombay

• Built a line-following robot to pick scattered magnetic pebbles and deposit them in a deposition zone, through the national-level E-Yantra robotics competition.
• Path-planned using A* for shortest collection time.
• Augmented-reality visualization of pebbles, water level, and the crow (robot), updated live on each pick/drop.

Code Videos