Keval Doshi

Hello! I am an Applied Scientist II at Amazon Ring AI (since July 2025), where I work on vision-language models (VLMs) and LLM-based multimodal reasoning for video understanding and object detection. Previously, I was an Applied Scientist II at Amazon Prime Video (June 2022 - June 2025).

I earned my Ph.D. in Electrical Engineering at the University of South Florida in the Secure & Intelligent Systems Lab, working under the direction of my advisor, Dr. Yasin Yilmaz. During my Ph.D., my research interests included computer vision, continual learning, and statistical anomaly detection. Previously, I earned an M.S. in Electrical Engineering from USF in 2018.

news

Jul 2025	Joined Amazon Ring AI as an Applied Scientist II.
Nov 2023	Won the Outstanding Dissertation Award from USF Graduate Studies.
Jun 2023	A multimodal benchmark and improved architecture for zero shot learning was accepted to WACV 2024.
Mar 2023	Dynamic inference with grounding based vision and language models was accepted to CVPR 2023.
Feb 2022	I will be joining Amazon Science as an Applied Scientist.
Oct 2021	Rethinking Video Anomaly Detection - A Continual Learning Approach was accepted to WACV 2022.
Oct 2021	A Modular and Unified Framework for Detecting and Localizing Video Anomalies was accepted to WACV 2022.
Jun 2021	First place in the CVPR 2021 Continual Learning Challenge - Reinforcement Learning Track. Leaderboard.
May 2021	Recognized as an “Outstanding Reviewer” for CVPR 2021.
Apr 2021	An Efficient Approach for Anomaly Detection in Traffic Videos was accepted to CVPR 2021 - AI City Workshop.
Jan 2021	Incoming Summer Intern at Nokia Bell Labs.

awards

Outstanding Dissertation Award (USF Graduate Studies)
1st Place, CVPR 2021 Continual Learning Challenge (Reinforcement Learning Track)
Outstanding Reviewer, CVPR 2021
Winner, NIST Automated Streams Analysis for Public Safety (2020)
2nd Place, NVIDIA AI CITY Challenge (CVPR 2020)