AI Engineer & Researcher
Zerui Wang
Building explainable & trustworthy AI systems that humans can understand and trust. Ph.D. in Computer Engineering from Concordia University.
Research Focus
Explainable AI
Making AI decisions transparent and interpretable for trust and compliance
Video Understanding
Transformer models for real-time video analysis with spatio-temporal attention
AI Security
Adversarial attacks and defenses for robust neural network systems
Cloud AI Services
Production XAI deployment on Azure, GCP, and AWS platforms
Publication Venues
All publications →Featured Projects
View all →
STAA - Spatio-Temporal Attention Attribution
5 citationsIEEE Access 2025 (Q1, IF: 3.6)
A real-time XAI method for interpreting video Transformer models. STAA extracts both spatial and temporal explanations simultaneously in a single forward pass, with less than 3% computational overhead. Achieves superior faithfulness (0.87) and monotonicity (0.91) scores on Kinetics-400 dataset with sub-150ms latency for real-time applications.
XAIport - Early XAI Adoption Framework
11 citationsICSE 2024 (A* Conference, ~20% acceptance)
A microservice-based framework for integrating Explainable AI into ML development pipelines. XAIport shifts explainability from post-hoc analysis to an integral development practice. Compatible with Azure Cognitive Services, Google Cloud Vertex AI, and AWS Rekognition. Improves both model performance and explanation stability.
Joint Spatio-Temporal Adversarial Attack
ACM TOMM 2025 (Q1, IF: 6.0)
A novel adversarial attack framework targeting video Transformer models through XAI-guided gradient perturbation. First method to simultaneously perturb spatial and temporal features. Achieves highest Attack Success Rate (ASR) on Kinetics-400, significantly exceeding V-BAD and sparse attacks. Reveals systematic security vulnerabilities in video transformers.
Latest Posts
View all →STAA: Real-Time Explanation for Video Transformers
Introducing STAA (Spatio-Temporal Attention Attribution), a novel method for real-time interpretation of video Transformer models. Published in IEEE Access 2025.
XAIport: Bringing Explainability to MLOps Pipelines
XAIport is a microservice framework for early adoption of Explainable AI in ML development. Published at ICSE 2024.
Interested in Collaboration?
I'm always open to discussing research collaborations, consulting opportunities, or just chatting about AI and explainability.