Nov 2023 – Feb 2024 · 3 mos · Washington DC-Baltimore Area

• Improved Multi‑Center Generalizability of GAN‑Based non‑FS to FS MR scan generation using Federated Learning with Pix2Pix architecture achieving 36.9% improvement in SSIM over FastMRI.
• Conducted EDA on the impact of temperature variations on health disparities using radiology examination metadata.

Jul 2023 – Nov 2023 · 4 mos · New York, New York, United States

• Built a U‑Net based approach to generate 3D images of contrast‑enhanced MRIs from non‑contrast scans orchestrated using PyTorch Lightning.
• Validated the approach on NYUMets database achieving 0.74 SSIM score.

Jan 2023 – Jul 2023 · 6 mos · New York, New York, United States

• Investigated improvement in video‐based tasks for neuroscience research through representation learning using part‐whole hierarchies in the GLOM model.
• Adapted the concept Deep Learning model’s implementation for images to video using PyTorch Lightning, TorchVision, and TorchVideo.

Aug 2022 – Dec 2022 · 4 mos · New York, New York, United States

• Performed Exploratory Data Analysis on Lung Adenocarcinoma cell scans to predict less common EGFR mutation.
• Preprocessed the high resolution whole slide image files (5GB per image) with OpenSlide.
• Tailored the registration process for pixel‑to‑pixel correspondence using Python, OpenCV, and scikit‑image.

Jul 2023 – Sep 2023 · 2 mos · Baltimore, Maryland, United States

• Conducted in‐depth analysis of social media interactions related to diabetes drugs (GLP‐1) to gain valuable insights into public perception.
• Constructed a web scraping pipeline for collecting and preprocessing tweets using Twitter API v2, Python, and Tweepy.
• Applied topic modeling using BERTopic, dimensionality reduction, and TF‐IDF scores to identify key themes.
• Performed sentiment analysis with RoBERTa to quantitatively assess the sentiments expressed in tweets.

Jan 2023 – May 2023 · 4 mos · New York, New York, United States · On-site

• DS-UA 301 Advanced Topics in ML/DL (Prof. Parijat Dube)
• Conducting weekly lab sessions
• Answering questions on Brightspace
• Holding office hours
• Working on course material

Sep 2022 – May 2023 · 8 mos

• CSCI-GA 2271 Computer Vision (Prof. Rob Fergus)
• CSCI-GA 3033 Introduction to Deep Learning Systems (Prof. Parijat Dube)
• CSCI-UA 310 Basic Algorithms (Prof. Marshall Ball)
• CSCI-UA 310 Basic Algorithms (Prof. Vladimir Podolskii)
• CSCI-UA 472 Artificial Intelligence (Prof. Ernest Davis)

May 2022 – Aug 2022 · 3 mos

• Developed stackable CNN ‐ Transformer blocks to build deeper Vision Transformer networks.
• Evaluated the framework on COCO dataset for object detection tasks.

Jun 2022 – Aug 2022 · 2 mos · Boston, Massachusetts, United States

• • Instrumented LeCAR Machine Learning‑based caching system on their Mission Uplink platform increasing cache‑hit ratio by 10% reducing network latency and enhancing internet connectivity for underserved communities.

Jun 2019 – Aug 2021 · 2 yrs 2 mos · Delhi, India

• Optimized denoising on iris and palmprints by reducing Non‑Local Means filter time by 75%, achieving PSNR up to 40.86 and 98.39% classification accuracy with a pretrained ResNet50 in PyTorch.
• Developed DeepCrypt, integrating CNNs with cryptographic hashing to enable secure cloud storage of biometric templates, achieving 99.56% authentication accuracy.
• Built a multimodal biometric authentication system using feature fusion leveraging iris and face modalities orchestrated on TensorFlow and Keras, achieving 99.8% accuracy while mitigating security limitations.

Jan 2018 – May 2018 · 4 mos · Delhi

• Designed a full‐stack web‐based solution to connect users with mental health specialists using HTML and CSS for frontend and PHP for backend.
• Improved accessibility and support for users by utilizing Skype API to set up meetings and send invites.
• Streamlined the dynamic RSS feed system to provide relevant articles on mental health based on user preferences and behavior, resulting in a 20% increase in user engagement.