Jun 2023 – Dec 2023 · 6 mos · New York, New York, United States · On-site

• I worked with the AudioBox team, and was able to build a new model for the task of text to music generation. Given a textual description of a piece of music, I designed a novel architecture to achieve SOTA performance in various generative music tasks.

Oct 2020 – Jul 2025 · 4 yrs 9 mos · Oxford, England, United Kingdom · Hybrid

• I worked on video understanding tasks, and trained models while using limited amount of manually labeled data. I published multiple papers on lip reading, British sign language recognition, co-speech gesture understanding. The underlying theme has been to learn to solve these tasks with weak supervision.

Jul 2019 – Nov 2020 · 1 yr 4 mos

• Recent work on accurately lip-syncing videos to any target speech has been accepted at ACM Multimedia 2020! Links to the demo video and code are attached!
• Our work on generating speech from only the lip movements has been accepted at CVPR 2020! Please watch the attached demo video! And check out the code, dataset, and the paper as well.

Jun 2018 – Jul 2019 · 1 yr 1 mo

• Worked on two projects in vision, deep learning, and multimodal understanding:
• 1. Making memes accessible to the visually impaired users by describing the meme face emotions using rich captions. The work is accepted as a full paper in ACM Multimedia 2019.
• 2. Translating talking face videos in a source language to a desired target language with complete lip synchronization. This work is also accepted as a full oral paper in ACM Multimedia 2019. Watch the demo video below!

May 2017 – Aug 2017 · 3 mos · Singapore

• Developed a novel architecture for OCT image segmentation of the Optic Nerve Head. The model achieved state of the art results in segmentation across multiple OCT machines.
• I also made a significant contribution to another ongoing project by developing tools to better understand the distinguishing features of healthy and glaucoma eyes using Class activation maps (CAMs). Fine-tuned a pretrained ResNet50 model to perform the classification, and highlighted the area of the OCT scan that the CNN focuses on to make its prediction.

Dec 2016 – Jan 2017 · 1 mo · Chennai, Tamil Nadu, India

• I worked on a Computer Vision problem under Prof. Dr. Anurag Mittal. Specifically, I worked on Object (foot) localisation and toe-tip detection in images of extreme variations in brightness, contrast and scale. I implemented and trained a Deep Convolutional Neural Network (CNN) from scratch on a small dataset with severe data augmentation and regularisation. I achieved excellent results approved by the Professor.

May 2016 – Jul 2016 · 2 mos · Delhi, India

• I worked under Dr. Naveen Garg, Professor, CS, IIT Delhi. I was accepted through the IASc Summer Research Fellowship programme (SRFP).
• I was engaged in developing a better approximation algorithm for the Capacitated K-Center Problem. I was able to come up with several novel ideas and explore various approaches during the research team's constant efforts to solve the problem.
• I also was involved in implementing and analyzing the accuracy of a randomized algorithm for the same problem. The algorithm was analyzed on several practical graph datasets and a detailed report was compiled. I was able to develop a keen taste for research during this period.

Dec 2015 – Dec 2015 · 0 mo · Mumbai

• Working as a backend Python Web Developer, I was responsible for establishing a better courier tracking system for the company. I had a wonderful experience in working with a well written Django(Python) legacy code. My work also involved integrating the existing system with third party APIs. The internship gave me an opportunity to work with a great tech team, thus teaching me the virtue of team coordination and quite a lot about real life software development.

Jul 2015 – May 2018 · 2 yrs 10 mos · NIT Trichy

• Delta Force is a premier community of enthusiastic developers working on projects and marching towards innovation in different areas of software engineering.

Feb 2015 – May 2018 · 3 yrs 3 mos · NIT Trichy

• I led an enthusiastic team of about thirty, to organize a university-level TEDx event. I was involved in bringing down speakers with brilliant ideas and curate their TEDx talk. I constantly encouraged ideation to ensure a holistic TED-like experience for the audience. Previously, as Head of Speakers team, I led a team of about six members and brought down a total of 13 distinguished speakers for 2 consecutive events.