Your Name

Zahra Gharaee, Ph.D.

200 University Ave West

Waterloo, ON, Canada N2L 3G1

    About me

  • 7+ years of experience in cross-functional teams, leading cross-industry projects across diverse machine learning domains.
  • Demonstrated successful leadership and management in multiple projects, achieving notable outcomes.
  • Expertise in building, debugging, and deploying resilient machine learning and deep learning models.
  • Developed scalable data solutions using TensorFlow and PyTorch, prioritizing usability and feasibility.
  • Released several datasets and code repositories, and published papers at prestigious venues such as NeurIPS and ICML.
  • Developed robust data pipelines for automated data ingestion and preprocessing, significantly improving data accessibility and reducing processing times.
  • Collaborated with multidisciplinary teams, effectively communicating complex concepts to non-technical stakeholders and ensuring alignment on project objectives and deliverables.
  • Actively involved in mentoring and supervising junior researchers on best practices for conducting projects, as well as writing and presenting research outcomes.
  • Served as a reviewer for multiple high-impact journals and conferences, providing constructive feedback to enhance the quality of research submissions and ensuring rigorous academic standards.

Expertise

  • Machine learning
  • Artificial Intelligence
  • Biotechnology
  • Neural Networks
  • Deep Learning
  • Computer Vision
  • Pattern Recognition
  • Feature Engineering
  • Big Data Analytics
  • Statistical Processing
  • Supervised Learning
  • Classification
  • Unsupervised Learning
  • Clustering
  • Graph Convolutional Networks
  • Causal Inference
  • Generative AI
  • Variational Autoencoders
  • 3D Shape Reconstruction

Experience

Research Associate

11/2024 - Current

Vision and Image Processing Lab (VIP)

University of Waterloo, Waterloo, Canada

Research Fellow

02/2022 - 11/2024

Vision and Image Processing Lab (VIP)

University of Waterloo, Waterloo, Canada

  • Releasing and Benchmarking Multimodal Datasets for ML Applications
    • Collaborative projects across multiple partners.
    • Led data science projects involving data structures, data solutions, data migration, and data governance.
    • Generated metadata while performing big data analytics and processing.
    • Leveraged vision software and libraries for machine learning applications, including image classification.
    • Released multi-modal datasets; over 1 million, and over 5 million on research and data sharing platforms (e.g., Zenodo, Kaggle, and Hugging Face).
    • Released two code repositories on version control systems (e.g., Git) and gained familiarity with AWS.
    • Benchmarked image classification against baselines (ResNet50 and Transformers) by leveraging transfer learning.
    • Fine-tuned backbone models and achieved over 90% accuracy (Micro-F1, Macro-F1) on large-scale datasets.
    • Published articles in NeurIPS 2023 and NeurIPS 2024.
    • For more details, please visit BIOSCAN-1M and BIOSCAN-5M
  • Image and Video Processing
    • In collaboration with the Microsoft Media Group.
    • Authored a paper on video object's relationship detection, which enhanced detection performance by about 4% (IEEE 2023).
    • Mentored research project about applying Transformers on super-pixels for efficient salient object detection (SOD).
    • For more details, please visit MoE-VRD and SuperFormer
  • Causality
    • In collaboration with the Department of ECE University of Waterloo.
    • Mentored a project on generative causal inference, which enhanced domain generalization by about 8.8%.
    • Published an article in ICML 2023.
    • For more details, please visit Causality

Postdoc

09/2018 - 02/2022

Computer Vision Lab (CVL)

Linköping University, Linköping, Sweden

  • Autonomous driving from virtual to the real-world
    • In collaboration with the WASP and SCANIA
    • Managed projects on road network graph learning and reinforcement learning for autonomous driving.
    • Developed, debugged, and executed ML experiments in cloud, parallel computing, and bash environments.
    • Enhanced graph representation learning for unsupervised transductive by 2% and supervised inductive by 10%.
    • Collected, curated, and released geospatial data from OpenStreetMap (OSMnx) for 18 Swedish cities.
    • Released two code repositories on Git.
    • Published articles in Pattern Recognition 2021 and ICPR 2021.
    • For more details, please visit BRL-VBVC and GAIN
  • Single-image reconstruction of 3D shape, texture, and camera pose
    • Developed and implemented predictive models for self-supervised 3D shape estimation.
    • Enhanced Mean-IoU about 8%, and reduced 3D-Angular-Error of the pose about 5° for the CUB dataset.
    • Released code repository on Git.
    • For more details, please visit SLOPE-KP

Projects

BIOSCAN-1M

BIOSCAN-1M

About: Introducing, studying, and benchmarking the BIOSCAN-1M Insect dataset.

Keywords: Biotechnology, Data management, Big data analytics, Classification, Transfer Learning.

BIOSCAN-5M

BIOSCAN-5M

About: Introducing, studying, and benchmarking the BIOSCAN-5M multimodal dataset.

Keywords: Biotechnology, Data management, Big data analytics, Statistical processing, Feature engineering.

SuperFormer

SuperFormer

About: A Transformer-based model that leverages superpixels for efficient Salient Object Detection (SOD).

Keywords: Computer vision, Salient object detection, Efficient model, Fine-tuning.

SLOPE_KP

SLOPE-KP

About: An approach to self-supervised learning of object pose estimation by keypoints prediction.

Keywords: Computer vision, 3D shape reconstruction, Pose estimation, 3D rotation.

MoE-VRD

MoE-VRD

About: An approach to video relationship detection using mixture of experts.

Keywords: Computer Vision, Video relationship detection, Mixture of experts.

GAIN

GAIN

About: An approach to graph representation learning.

Keywords: Graph convolutional neural networks, Representation learning, Geospatial data analysis.

Causality

Causality

About: Two approaches to generative and implicit causal representation learning.

Keywords: Causal inference, Generative models, Interventions, Variational autoencoders.

BRL_VBVC

Autonomous Driving

About: An approach to Bayesian Reinforcement Learning of vision-based vehicular control.

Keywords: Autonomous systems, Reinforcement learning, Simulated environment.

Education

Ph.D. in Cognitive Science

03/2014 - 05/2018

Thesis: Action in mind: A neural network approach to action recognition and segmentation.

Lund University Cognitive Science (LUCS)

Lund University, Sweden

M.Sc. in Mechatronics

09/2009 - 02/2012

Thesis: Attention control learning in Decision Space Using State Estimation.

Advanced Process, Automation and Control Research Group (APAC)

K.N. Toosi University of Technology, Iran

B.Sc. in Electrical Engineering

09/2005 - 09/2009

Thesis: Design and implementation of a MIMO controller for a quadruple tank system.

Advanced Process, Automation and Control Research Group (APAC)

K.N. Toosi University of Technology, Iran

Public Speaking

Computational Entomology Webinar III: Processing liquid samples

06/2024

About: Computer vision and robotic tools to automatically process insect specimens stored in liquid samples.

Vancouver, Canada

GEO BON Global Conference: Monitoring Biodiversity for Action

10/2023

About: AI for Insect Monitoring.

Montreal, Canada

LiU Game Conference

11/2018

About: Computer games, visualization and digital experiences.

Linköping, Sweden

Skills

Selected Publications