Shehreen Azad

Applied Scientist II, Ring · Amazon | PhD Candidate, CRCV · UCF

I am an Applied Scientist II at Ring, Amazon, where I work on multimodal video understanding for real-world streaming systems.

I am also a final year PhD candidate at the Center for Research in Computer Vision (CRCV), University of Central Florida, advised by Dr Yogesh Singh Rawat.

My research sits at the intersection of deep learning and computer vision, focused on long-form and streaming video understanding and multimodal reasoning.

Email Google Scholar Resume GitHub LinkedIn

News

Sep 2026 Joined Ring, Amazon as an Applied Scientist II.

Sep 2026 One co-author benchmark paper accepted at NeurIPS 2026.

Jul 2026 One co-author paper accepted at ECCV 2026.

May 2026 Started as an Applied Scientist Intern at Ring, Amazon.

Feb 2026 One first-author paper accepted at CVPR 2026.

Jun 2025 One first-author paper accepted at ICCV 2025 as a Highlight.

Mar 2025 One benchmark paper accepted at the CVPR MMFM Workshop 2025.

Feb 2025 One first-author paper accepted at CVPR 2025.

Mar 2024 Two benchmark papers accepted at the CVPR MMFM Workshop 2024.

Feb 2024 One first-author paper accepted at CVPR 2024.

Selected Publications

Reverse chronological. A complete list is on my Google Scholar.

Remembering What Matters: Balanced Memory for Ultra-Long Streaming Videos

Shehreen Azad, Mo Zhou, Shan Lu, Wei Wang, Noranart Vesdapunt, Sowndarya Sundar, Jim Thomas, Kah Kuen Fu

ICLR 2027 Under review

A streaming video model with a balanced memory mechanism that effectively processes week-long streaming video without forgetting early information.

StreamReady: Learning What to Answer and When in Long Streaming Videos

Shehreen Azad, Vibhav Vineet, Yogesh Singh Rawat

CVPR 2026

A memory-augmented, task-agnostic multimodal framework that gives Multimodal Large Language Models real-time, continuous understanding of long-form streaming video — learning both what to answer and when to answer under a streaming budget.

Project Page Paper arXiv Code Data

DisenQ: Disentangling Q-Former for Activity-Biometrics

Shehreen Azad, Yogesh Singh Rawat

ICCV 2025 Highlight Patent granted

A disentanglement-based multimodal architecture for robust activity-aware person identification, using language-guided, motion-centric temporal modeling to separate identity from appearance.

Project Page Paper arXiv Code

HierarQ: Task-Aware Hierarchical Q-Former for Enhanced Video Understanding

Shehreen Azad, Vibhav Vineet, Yogesh Singh Rawat

CVPR 2025

A task-aware, memory-augmented hierarchical framework that equips Multimodal Large Language Models for efficient understanding of arbitrarily long-form video.

Project Page Paper arXiv Code

Activity-Biometrics: Person Identification from Daily Activities

Shehreen Azad, Yogesh Singh Rawat

CVPR 2024 Patent pending

Introduced the task of activity-biometrics, scaling gait-based person re-identification to activity-aware re-identification, with a disentanglement-based framework linking actions to biometric traits.

Paper arXiv Code

Experience

Sep 2026 – Present

Applied Scientist II

Ring, Amazon · Bellevue, WA

Multimodal video understanding for production streaming systems, spanning long-form temporal reasoning, efficient modeling under tight compute budgets, and evaluation of multimodal large language models on real-world video.

May 2026 – Jul 2026

Applied Scientist Intern

Ring, Amazon · Bellevue, WA

Built a balanced-memory streaming architecture for ultra-long video understanding, retaining early context across week-long streams. Led to a first-author submission to ICLR 2027.

Education

Aug 2023 – Dec 2026

PhD, Computer Science Expected

University of Central Florida · GPA 3.92 / 4.0

Advised by Dr Yogesh Singh Rawat. Committee: Dr Yogesh Singh Rawat, Dr Mubarak Shah, Dr Chen Chen, Dr Soheil Sabri.

Aug 2022 – Jul 2023

MS, Computer Science

University of Central Florida

Patents

Shehreen Azad, Yogesh Singh Rawat. Activity-based person identification using biometric disentanglement. Granted
US 12,482,255 B1 · 2025

Shehreen Azad, Yogesh Singh Rawat. Silhouette-guided feature distillation for activity-aware person identification. Pending
US 2026/0057661 A1 · filed 2025

Awards

3× Diversity, Equity and Inclusion Award CVPR'24, ICCV'25, CVPR'26

3× UCF CS Ranking Incentive Award '24, '25, '26

2× UCF Presentation Fellowship '24, '25

2^nd place, IARPA BRIAR — Biometric Recognition and Identification at Altitude and Range

UCF ORCGS Doctoral Fellowship 2023–2024

Service

Reviewer, CVPR 2025, 2026

Reviewer, ICCV / ECCV 2025, 2026

Reviewer, NeurIPS 2025, 2026

Reviewer, ICLR 2026, 2027

Reviewer, AAAI 2026, 2027

Reviewer, BMVC 2025, 2026

Reviewer, ICPR 2024, 2026

Teaching Assistant, UCF CS Computer Graphics, Ray Tracing · 2022–2023

Treasurer, Bangladesh Student Association at UCF 2023–2024