Projects

Academic Projects

  • Spot the Door: Developed a Flask server to enable a low-latency seamless interaction between VR and quadrupedal robot and engineered a precise world-to-pixel coordinate conversion algorithm, enabling accurate manipulation of Spot's actions in 3D
  • InterTwine: Augmented the functionality of the browser-cache based open-source story-telling tool, Twine to enable real-time collaboration
  • Accuracy-Privacy Tradeoffs in SOTA Weak-Supervision Models: In this paper, we added a layer of differential privacy to two existing weak-supervision frameworks, Ratner et al. [2017] and Karamanolakis et al. [2021] and analyze the privacy-accuracy trade-offs with the added layer of privacy
  • Private Matching: A survey of private set intersection protocols over the decades for future compute
  • TrainDy: a full stack H&M fashion recommendation system application with using a combination of techniques such as collaborative and content-based filtering and a variety of machine learning models such as k-means clustering and conditional variational autoencoders.
  • viVID-GAN: implementation of SRGAN: a GAN model architecture for image super-resolution that produced images of 4x perceptual quality from their low-resolution counterparts. We modified the loss function to generalize the model across domains and extended the idea to videos by upsampling each frame.
  • Capstone Project: studied and implemented privacy-preserving linear regression in two frameworks, namely, SecureML (2017) in the 2-server setting and BLAZE (2020) in the 3-server setting and proposed optimisations in the multiplication protocols and truncation.
  • BlockSim-Net: a comprehensive simulator in Python for simulating attacks and attack-mitigation protocols in Bitcoin and Ethereum.
  • HyperFunds: a distributed full stack application on Hyperledger Fabric that enables faculty at Ashoka University to reliably spend their research funds with approval from the right administrators using Python and Javascript.
  • Neural Machine Translation: an LSTM Encoder-Decoder based model for machine translation from English to Hindi with Attention mechanism in Python.
  • Yao's garbled circuits: implementation of RSA-based oblivious transfer and Yao's garbled circuits in C.