About
AI research scientist building evaluation infrastructure, perceptual metrics, and adversarial failure harnesses for transformer-based generative and multimodal models. Ph.D. from Concordia University (2025); 12 publications, first-author at ICSE, IEEE Transactions, and ACM Transactions. I designed the first joint spatio-temporal adversarial attack and defense on video transformers, and the first real-time interpretability method for transformer-based video models. I've built and deployed cloud-agnostic XAI / evaluation frameworks across Azure, GCP, and AWS, and served 40+ peer reviews at IEEE Transactions & AAAI. Currently shipping production multimodal generative pipelines at Maket.AI.
Experience 2019 – present · evaluation, eval infrastructure, production GenAI
Maket.AI
· Generative AI Engineer · Montreal
Aug 2025 – Present
Pipelines
Production multimodal generation — multi-stage LLM orchestration over floor-plan and design-brief inputs to drive architectural design reasoning, 3D visualization, and iterative AI-chat design loops.
3D Vision
SAM3D segmentation; Gaussian Splatting / NeRF reconstruction from phone video for in-app 3D scene capture and visualization.
AI Chat
Real-time, context-aware vision & language responses driving iterative design refinement; robust prompt orchestration for reliable outputs across multi-stage generation workflows.
Concordia University
· Doctoral Researcher · Montreal
May 2020 – Jun 2025
STAA
Novel perceptual metrics for video transformers — first single-pass spatio-temporal attention attribution, defining faithfulness and monotonicity fast enough to run as eval-as-CI inside the training loop. Validated on standard large-scale video benchmarks. IEEE Access 2025.
Red-team
Adversarial red-teaming & regression harness — first joint spatio-temporal adversarial attack on video transformers via XAI-guided gradient perturbation, systematically surfacing failure modes at scale. Reversed, the same machinery becomes adversarial-aware training that materially hardens the model. ACM TOMM 2025.
XAIport
Scalable automated evaluation infrastructure — microservice framework turning evaluation from a manual notebook into an automated MLOps stage; OpenAPI 3.0 contracts; cloud-agnostic orchestration of heterogeneous backends; regression testing, dashboards, explanation-drift monitoring, validated at scale. ICSE 2024; IEEE SSE 2025.
Multi-cloud
Cross-platform comparative evaluation across Azure Cognitive Services, GCP Vertex AI, and AWS Rekognition behind unified REST contracts; containerized microservices; reproducible benchmark harness. IEEE TCC 2024.
Quantum
· AI Engineer · Montreal
Apr 2023 – Aug 2025
Diligence
Cross-verified funding amounts, dates, lead investors, and valuation deltas across Crunchbase, PitchBook, and CB Insights for AI / infrastructure startups.
Eval
Tech-stack-anchored startup scoring along 5 axes (data moat, representation quality, retrieval architecture, evaluation rigor, online monitoring) — the same evaluation framework from my XAI research, reapplied to investment diligence and a bilingual LLM research pipeline with inline citations.
Université de Montréal
· Research Associate · Montreal
Sep 2019 – Mar 2021
Solvers
Built finite-volume, finite-element, Runge–Kutta, and spectral solvers for Navier–Stokes, advection–diffusion, and control-system equations in Python / MATLAB / COMSOL; numerical simulation of fluid transport and multi-physics coupling, validated against analytical benchmarks.
Selected Publications 12 total · first-author at ICSE, IEEE Trans., ACM Trans.
Joint Spatio-temporal Adversarial Attacks on Video Transformer Models Through XAI-guided Gradient Perturbation2025
Wang, Z., Liu, Y. · ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM)
STAA: Spatio-Temporal Attention Attribution for Real-Time Interpreting Transformer-Based Video Models2025
Wang, Z., Liu, Y. · IEEE Access, vol. 13, pp. 101647–101661
An Open API Architecture to Discover the Trustworthy Explanation of Cloud AI Services2024
Wang, Z., Liu, Y., Huang, J. · IEEE Transactions on Cloud Computing, vol. 12, no. 2, pp. 762–776
XAIport: A Service Framework for the Early Adoption of XAI in AI Model Development2024
Wang, Z., Liu, Y., Thiruselvi, A.A., Hamou-Lhadj, W. · 46th IEEE/ACM Int. Conf. on Software Engineering (ICSE), Lisbon, pp. 67–71
Cloud-Based XAI Services for Assessing Open Repository Models Under Adversarial Attacks2024
Wang, Z., Liu, Y. · IEEE Int. Conf. on Software Services Engineering (SSE), pp. 141–152
XAIpipeline: Automated Orchestration of Explainable AI Service for Cloud AI and Open-Source Models2025
Wang, Z., Liu, Y. · IEEE Int. Conf. on Software Services Engineering (SSE)
Spatio-temporal Explanation for Adversarial-Aware Cloud Vision Services2025
Wang, Z., Liu, Y. · IEEE Computer Software and Applications Conference (COMPSAC)
The Role of Provenance Modeling in Tracing and Reproducing Explainable AI Pipelines2025
Wang, Z., Liu, Y. · Int. Conf. on Computational Science and Computational Intelligence (CSCE)
The Analysis and Development of an XAI Process on Feature Contribution Explanation2022
Huang, J.*, Wang, Z.*, Li, D., Liu, Y. · IEEE Int. Conf. on Big Data, Osaka, pp. 5039–5048
A Trustworthy View on Explainable Artificial Intelligence Method Evaluation2023
Li, D., Liu, Y., Huang, J., Wang, Z. · IEEE Computer, vol. 56, no. 4, pp. 50–60
Linking Team-level and Organization-level Governance in MLOps through Explainable AI and Responsible AI Connector2023
Neghawi, E., Wang, Z., Huang, J., Liu, Y. · IEEE Computer Software and Applications Conference (COMPSAC), pp. 1223–1230
VideoXAI: A Hybrid Architecture for Explainable AI Pipelines of Robust Video Classification2025
Kondal, A.S., Ghataura, R.S., Liu, Y., Wang, Z. · IEEE Int. Conf. on Big Data
Full list on Google Scholar →
Core Skills
Evaluation
Perceptual metrics, faithfulness & monotonicity, adversarial red-teaming, regression & explanation-drift monitoring, eval-as-CI, benchmark-harness design across video / image / text / audio / tabular.
Multimodal & GenAI
Video & image transformers, LLM orchestration, multi-stage generation pipelines, real-time vision-language; production generative AI shipping.
3D Vision
SAM3D segmentation, Gaussian Splatting, NeRF reconstruction, phone-video scene capture and in-app visualization.
Infra & MLOps
Microservices, OpenAPI 3.0, cloud-agnostic orchestration across Azure / GCP / AWS; containerization; dashboards, alerting, and large-scale evaluation pipelines.
Foundations
PDE & numerical solvers (FVM / FEM / Runge–Kutta / spectral), control theory, statistics; Python, PyTorch, MATLAB, COMSOL.
Education
Concordia University · Ph.D., Computer Engineering · explainability & adversarial robustness for video transformers
2020 – 2025
Technical University Dortmund · M.Sc., Process Systems Engineering · modeling, PDE, control theory, computational simulation
2014 – 2018
China University of Mining & Technology · B.Sc., Chemical Engineering
2010 – 2014