Hi! My name is Nikhil!

I’m a PhD student at Georgia Tech, advised by Alexey Tumanov. Previously, I built AI platforms for document understanding at Bevaya AI.

I earned a Masters from NYU’s Courant Institute, where I was a part of the CILVR group working with Lerrel Pinto and Rob Fergus on skill learning and offline RL for imperfect information games.

Before NYU, I explored novel techniques for conversational question answering with Danish Contractor and team at IBM Research India. Subsequently, at Flixstock, I focused on GAN based approaches to photorealistic image generation.

I also spent some time at Microsoft Research India working with Navin Goyal on translating code to natural language. I hold a bachelor’s degree in Electrical Engineering from BITS Pilani. For more information, you can have a look at my resume or reach out to me through email.

Publications

[Abstract] Page Stream Segmentation with LLMs: Challenges and Applications in Insurance Document Automation
Hunter Heidenreich, Ratish Dalvi, Nikhil Verma, Yosheb Getachew
Paper published at COLING 2025 (Industry Track)

[Abstract] Neural Conversational QA: Learning to Reason vs Exploiting Patterns
Nikhil Verma, Abhishek Sharma, Dhiraj Madan, Danish Contractor, Harshit Kumar, Sachindra Joshi
Paper published at EMNLP 2020

[US 16/892805] Generating Dialog System Workspaces
Danish Contractor, Nikhil Verma, Harshit Kumar and Sachindra Joshi
Patent filed with the US Patent and Trademark Office

Open Source

[GitHub] SHIS: Simple HTTP Image Server
A drop-in replacement for python -m http.server, albeit for images.