May 2024 – Aug 2024 · 3 mos · San Francisco, California, United States · On-site

• Factored approach of first retrieving relevant moments in long video and then answering questions. Benchmarked approach with 4 large VLMs (proprietary and open) on 4 long form video QA datasets.

Jun 2022 – Jul 2022 · 1 mo · Tokyo, Japan

• Worked on explainable AI. Developed 3 well known methods as part of an under development library with proper testing and unified API. Reproduced Attention Branch network with different backbones on Sony's neural network console (NNC, a GUI for building neural nets).

May 2022 – Oct 2022 · 5 mos · Delhi, India

Jul 2021 – Apr 2022 · 9 mos · Delhi, India

Jun 2021 – Jun 2022 · 1 yr · Delhi, India

Oct 2020 – May 2021 · 7 mos · Delhi, India

Oct 2020 – May 2021 · 7 mos · Delhi, India

May 2021 – Aug 2021 · 3 mos · Sheffield, England, United Kingdom

• Worked on maintenance and enhancements of PyKale, a library providing unified API design for various methods in Domain Adaptation. Co-Authored the paper proposing PyKale, submitted to CIKM 2022. Integrated TA3N, State of the Art model in Unsupervised Action Domain Adaptation, into the library. Explored Partial Domain Adaptation and worked on novel ideas addressing Negative Transfer in videos.

Nov 2020 – Dec 2020 · 1 mo · Bengaluru, Karnataka, India

• Worked on obtaining clear and accurate boundaries for Indian farms using the Sentinel 2 Satellites data. Built a complete pipeline for extracting boundary vectors using Mask-RCNN for instance segmentation. Explored basics of remote sensing, to effectively utilise multiple wavelength information for better results.

Apr 2020 – Jul 2020 · 3 mos · Pune, Maharashtra, India

• Leveraged various Deep Learning and Computer Vision techniques for the unsolved problem of automated box localisation and counting in Bulk storage warehouses for the drone-based automated inventory management system.