Shravan Nayak

prof_pic.jpg

Hi! I’m Shravan Nayak. I am currently pursuing my research master’s (and soon a PhD) in computer science at Mila and the University of Montreal supervised by Prof. Aishwarya Agrawal. My research focuses on cultural understanding in AI systems—how AI can recognize, respect, and adapt to different cultural perspectives. I believe this is crucial not only for creating AI that is inclusive, safe, and fair but also for tackling broader challenges such as pluralistic alignment and improving its ability to handle complex real-world scenarios (robustness).

Currently, I am particularly interested in:

  1. Understanding how diverse human perspectives shape AI systems and developing methods to integrate these perspectives effectively while ensuring robust evaluation in real-world scenarios.
  2. Exploring how cultural understanding can serve as a framework to tackle key challenges like pluralistic alignment, multilingualism, and long-tailed robustness.

I also spend part of my time at ServiceNow Research, working on Multimodal Agents for Desktop Understanding. My work involves creating benchmarks to comprehensively evaluate AI agents’ ability to perform tasks in desktop environments and developing methods to improve their adaptability.

Prior to starting my graduate studies, I spent two years at Microsoft, contributing to the PowerPoint mobile team. If you’ve ever used PowerPoint on your iPad or mobile phone, chances are you’ve encountered my work, even if unknowingly! Alongside this, I’ve been deeply involved in artificial intelligence research for over six years. I have worked on reinforcement learning for energy management in microgrid networks at IISc, dubbing at LT Lab in Hamburg, and machine translation for low-resource languages at the University of Toronto.

In my free time, I love to cook and share my experiences on my blog. I also have a deep interest in history, culture, and archaeology, as I find it fascinating to explore the diverse ways of life and the events that have shaped our world. Currently, I am reading Early Indians by Tony Joseph and The Golden Road by William Dalrymple. One day, I hope to study the history of the entire world in depth.

News

Feb 13, 2025 Received ESP AI Scholarship of 5000 CAD from University of Montreal :)
Jan 31, 2025 BigDocs: An Open and Permissively-Licensed Dataset for Training Multimodal Models on Document and Code Tasks got accepted at ICLR 2025.
Nov 15, 2024 “MID-Space: Aligning Diverse Communities’ Needs to Inclusive Public Spaces” got an oral presentation at NeurIPS 2024 Pluralistic Alignment Workshop and “BigDocs: An Open and Permissively-Licensed Dataset for Training Multimodal Models on Document and Code Tasks” got accepted at RBFM Workshop at NeurIPS 2024.
Sep 19, 2024 Two papers accepted at EMNLP!! “Benchmarking Vision Language Models for Cultural Understanding” got accepted to Main and “Improving Adversarial Robustness in Vision-Language Models with Architecture and Prompt Design” got accepted to Findings. See you in Miami :)
Aug 26, 2024 Started as a Visiting Researcher at ServiceNow Research working on building the next generation of foundation models and datasets for document, web and GUI understanding.