Apoorv Umang Saxena

I am a research scientist at Adobe Research, Bangalore with a research focus on NLP and multi-modal machine learning.

Previously I received my PhD from the Department of Computational and Data Sciences (CDS) at the Indian Institute of Science, Bangalore.

Prior to graduate school, I received my Bachelor of Engineering (BE Hons.) in Computer Science from Birla Institute of Technology and Science, Pilani (BITS Pilani) in 2015.

Research

I am broadly interested in generative AI (LLMs, computer vision) and their applications. In my current role at Adobe, my research mainly focuses on enhancing the utility of LLMs for document-grounded tasks, such as summarization and question answering. This includes optimizing inference speed for document-centric tasks and developing methods to attribute LLM outputs back to the documents. I am also interested in design generation applications, specifically with (V)LLMs.

In my post-graduate work, I had a specific focus on knowledge graphs (KG) and question answering. In this area, I have tried to improve QA performance by combining pre-trained language models with KG embeddings. I have also worked on making KG embeddings more scalable and easier to apply to downstream tasks. During my undergraduate studies I worked on humanoid robotics and computer vision.

Research publications

Saxena A. Prompt Lookup Decoding. Developed a method to speed-up LLM decoding, integrated in transformers and vLLM.

Saxena A., Kochsiek A. & Gemulla R. “Sequence-to-Sequence Knowledge Graph Completion and Question Answering”. Accepted to the 2022 Annual Conference of the Association for Computational Linguistics (ACL 2022).

Saxena A., Chakrabarti S. & Talukdar P. “Question Answering Over Temporal Knowledge Graphs”. Accepted to the 2021 Annual Conference of the Association for Computational Linguistics (ACL 2021).

Saxena A., Tripathi A. & Talukdar P. “Improving Multi-hop Question Answering over Knowledge Graphs using Knowledge Base Embeddings”. Accepted to the 2020 Annual Conference of the Association for Computational Linguistics (ACL 2020).

A. Gautam, A. U. Saxena, P. Mall and S. Mohan, “Positioning multiple mobile robots for geometric pattern formation: An empirical analysis,” 2014 Seventh International Conference on Contemporary Computing (IC3), Noida, India, 2014, pp. .

Work Experience

Adobe Research, Bangalore

Research Scientist — June 2022 - Present

IBM Research, Bangalore

Research Intern — Summer 2019

Google, Hyderabad

Software Engineer, Tools and Infrastructure — May 2016 - Mar 2017

PayPal, Chennai

Software Engineer 1 — July 2015 - November 2015

Side Projects

3D Printing Store: Everything 3D

Get custom-designed products, including personalized fridge magnets, desk signs, house nameplates and more! Order your own unique items via the store or follow the latest creations on our Instagram :)

Store: everything-3d.mini.store
Instagram: everything_3d_india

Podcast transcription and highlight generation

I used openai-whisper (along with pyannote and flan-t5) to automatically generate transcripts and highlights for all episodes of The Seen and The Unseen podcast by Amit Varma. These can be browsed at https://apoorvumang.github.io/seenunseen