Santosh Patapati

Email: santosh[at]cyrionlabs[dot]org          

Google Scholar      Github      Resume      LinkedIn

I am currently a research intern at the University of Washington, and a Machine Learning Researcher and Co-Founder at Cyrion Labs. My work focuses on human-computer interaction, multimodal learning, visual-language models, and the semantic web.

In industry, I have worked on developing and scaling AI technology in roles ranging from Machine Learning Engineer to Full-Stack Developer in startups and established companies.

April 2025: One Paper (A General-Purpose Framework for Real-Time Adaptive Multimodal Embodied Conversational Agents) Accepted to Show & Tell @ Interspeech

April 2025: One Paper (PhysNav-DG: A Novel Adaptive Framework for Robust VLM-Sensor Fusion in Navigation Applications) Accepted to DG-EBF @ IEEE CVPR 2025

March 2025: Second Demo (VIZ: Virtual & Physical Navigation System for the Visually Impaired) Accepted to IEEE CVPR 2025 as a Demo

March 2025: One Demo (GenECA: A Generalizable Framework for Real-Time Multimodal Embodied Conversational Agents) Accepted to IEEE CVPR 2025 as a Demo

March 2025: One Preprint Paper Posted on arXiv

October 2024: Second Oral Presentation at IEEE MIT URTC

October 2024: First Poster Presentation at IEEE MIT URTC

September 2024: Poster Presentation at the Rice Ken Kennedy AI in Health Conference

August 2024: Research featured on The AI Timeline

July 2024: First Preprint Paper Posted on arXiv

June 2024: Joined as a research assistant at the SMU BAST Lab

May 2024: Began an internship on Nanotech x AI at the UW Department of Electrical & Computer Engineering

April 2024: First Poster Presentation at The Future of Biology Conference

Addresses the challenges faced by the visually impaired by utilizing generative AI to mimic human behavior for complex digital tasks and physical navigation.

Introduces a robust framework for multimodal interactions with embodied conversational agents, emphasizing emotion-sensitive interaction.

Presents a novel voice-controlled navigation agent using a ReAct-inspired architecture, offering improved accessibility for the visually impaired.

Presents a novel multimodal framework for automated MDD prognosis from clinical interview recordings. Proposes pipeline for real-world clinican use.

The website template was adapted from Jon Barron.