Bharat Runwal

Bharat Runwal

Undergrad @ IIT Delhi

IIT Delhi


Paper Reading


This website is no longer maintained Please Click Here to get to updated website, Thanks.

  • Computer Vision
  • NLP
  • Optimization Theory
  • Adversarial Robustness
  • B.Tech in Electrical Engineering, 2022

    Indian Institute of Technology(IIT) , Delhi


Deep Learning Researcher
July 2021 – Present Remote
Currently Working on theoretical Deep learning and applied computer vision.
Research Intern
June 2021 – August 2021 Remote
Worked on Quantization(Mainly 8-bit) of models using Quantization Aware Training for Object Detection and classification. Also worked on analyzing the degradation layer explicitly during the QAT. As a part of my work i also presented survey on application of Graph Neural Networks(GNNs) in industrial domain ex. Recommendation systems and why robustness is crucial for the deployed models for security level applications.
Junior Machine Learning Engineer
June 2021 – August 2021 Remote
I worked on the project “Helping People with Visual Impairment to Easily Use Buses through Computer Vision”,In collaboration with RenewSenses LTD, an Israeli company developing assistive technologies for people who are blind, this project involves assisting people with visual impairment in their experience of catching a bus. The outcome of this challenge will be directly tested on the community of people with visual impairment. I mainly contributed in team for Bus Detection and Tracking , where my goal was to track the “front of the bus” in the video real time.Used DeepSORT algorithm with the different backbone of YOLOv-4,YOLOv-5 etc. for tracking the “front of the bus” and speed of the bus.
NLP intern
May 2021 – June 2021 Remote
Worked on building a vernacular search engine for e-commerce applications with features like price tag detection from query, autocomplete,spell check.
Research Intern
HPI Potsdam,Germany
October 2020 – Present Remote
The project was on the Semantic Similarity based on Sense Embedding Induction,we used deconflation approach with additional constraint satisfacation enforcement which improved the semantic space of embeddings.Our proposed method gives competitive results on Simlex-999 and Simverb 3500 Datasets. (Code will be up in short amount of time after we finish our work)