Soroosh Shahtalebi

Principal Engineer

Specializing in Machine Learning, Generative AI, and Autonomous Agentic Systems

Soroosh Shahtalebi

About Me

As a Principal Engineer at Huawei Canada, I lead initiatives in autonomous agentic workflows and Large Language Models (LLMs). My journey from academic research to industry has included impactful contributions to AI, such as pioneering generative AI work at Unity Technologies, where I spearheaded the development of sound generation systems and human body modeling solutions.

My expertise spans both theoretical and applied AI, with high-impact publications in domain generalization and out-of-distribution learning. I’ve architected and deployed large-scale AI systems on cloud platforms like GCP and Azure, with strong MLOps and CI/CD practices. From explainable AI in healthcare to production-ready generative solutions, I focus on bridging theory and practice to tackle real-world challenges.

Experience

Principal Engineer

Huawei Canada

  • Leading research and development in autonomous agentic workflows and LLMs
  • Developing innovative solutions for AI-driven autonomous systems
  • Architecting and implementing large-scale AI systems

Senior Machine Learning Engineer

Unity Technologies

  • Led a team in developing a groundbreaking generative sound effect system from research concept to production deployment
  • Developed Generative AI solutions for materials, textures, and human body
  • Conducted research on Out-of-Distribution Generalization using synthetic data
  • Worked on the development of 3D object and scene reconstruction pipelines
  • Involved in the productionization and deployment of ML models at large-scale over GCP and Microsoft Azure

Postdoctoral Research Fellow

University of Toronto and Vector Institute for Artificial Intelligence

Project: Theory of Domain Generalization applied in Healthcare

Postdoctoral Research Fellow

Mila University of Montreal

Project: Theoretical perspectives on Out-of-Distribution Generalization

Part-time Consultant

MathWorks

Provided consultations to improve the Wavelet Toolbox of MATLAB software.

Part-time Technical Consultant

Ubico Labs

Advised on integrating ML into automated mailing systems.

CEO and Founder

Hooshmand Afzar Pars

Specialized in the design and production of Bio-feedback and Neuro-feedback devices.

Education

PhD in Information Systems Engineering

Concordia University, Montreal, QC, Canada (2016–2020)

With distinction, Outstanding grade. Project on Effective Application of Data-driven Learning Models for Assistive Technologies and Brain-computer Interfaces.

M.Sc. in Biomedical Engineering (minor in Bio-electrics)

K. N. Toosi University of Technology, Tehran, Iran (2013–2015)

With distinction. Project on Design and implementation of the processing and controlling circuit board of a blood cell counting machine.

Bachelor's in Electrical Engineering (major in Telecommunication)

Yazd University, Yazd, Iran (2009–2013)

Project on Design and production of a planar goniometer for joint angle measurement.

Selected Publications

Total Citations: 265

H-index: 10

2023

AnthroNet: Conditional Generation of Humans via Anthropometrics

F Picetti, S Deshpande, J Leban, S Shahtalebi, J Patel, P Jing, C Wang, ...

arXiv preprint arXiv:2309.03812 Citations: 3

2022

Predicting Fine-Tuning Performance with Probing

Z Zhu, S Shahtalebi, F Rudzicz

ACL 2022 Workshop BigScience Citations: 10

Ood-probe: A Neural Interpretation of Out-of-Domain Generalization

Z Zhu, S Shahtalebi, F Rudzicz

arXiv preprint arXiv:2208.12352 Citations: 3

A Remedy for Distributional Shifts through Expected Domain Translation

JC Gagnon-Audet, S Shahtalebi, F Rudzicz, I Rish

ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing Citations: 1

Out-of-Distribution Failure through the Lens of Labeling Mechanisms: An Information Theoretic Approach

S Shahtalebi, Z Zhu, F Rudzicz

ICML 2022: Workshop on Spurious Correlations, Invariance and Stability Citations: 1

2021

Sand-mask: An Enhanced Gradient Masking Strategy for the Discovery of Invariances in Domain Generalization

S Shahtalebi, JC Gagnon-Audet, T Laleh, M Faramarzi, K Ahuja, I Rish

ICML Workshop on Uncertainty and Robustness in Deep Learning (2021) Citations: 60

A Deep Explainable Artificial Intelligent Framework for Neurological Disorders Discrimination

S Shahtalebi, SF Atashzar, RV Patel, MS Jog, A Mohammadi

Scientific Reports, Nature Publishing Group Citations: 27

2020

PHTNet: Characterization and Deep Mining of Involuntary Pathological Hand Tremor using Recurrent Neural Network Models

S Shahtalebi, SF Atashzar, O Samotus, RV Patel, MS Jog, A Mohammadi

Scientific Reports, Nature Publishing Group Citations: 32

Siamese Neural Networks for EEG-based Brain-Computer Interfaces

S Shahtalebi, A Asif, A Mohammadi

2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society Citations: 20

Towards Effective Application of Data-driven Learning Models for Assistive Technologies and Brain-computer Interfaces

S Shahtalebi

Concordia University PhD Thesis

2019

HMFP-DBRNN: Real-time Hand Motion Filtering and Prediction via Deep Bidirectional RNN

S Shahtalebi, SF Atashzar, RV Patel, A Mohammadi

IEEE Robotics and Automation Letters Citations: 18

WAKE: Wavelet Decomposition Coupled with Adaptive Kalman Filtering for Pathological Tremor Extraction

S Shahtalebi, SF Atashzar, RV Patel, A Mohammadi

Biomedical Signal Processing and Control Citations: 29

2018

Bayesian Optimized Spectral Filters Coupled with Ternary ECOC for Single-trial EEG Classification

S Shahtalebi, A Mohammadi

IEEE Transactions on Neural Systems and Rehabilitation Engineering Citations: 30

A Bayesian Framework to Optimize Double Band Spectra Spatial Filters for Motor Imagery Classification

S Shahtalebi, A Mohammadi

2018 IEEE International Conference on Acoustics, Speech and Signal Processing Citations: 10

2017

Ternary ECOC Classifiers Coupled with Optimized Spatio-spectral Patterns for Multiclass Motor Imagery Classification

S Shahtalebi, A Mohammadi

IEEE International Conference on Systems, Man, and Cybernetics (SMC) Citations: 12

A Multi-rate and Auto-adjustable Wavelet Decomposition Framework for Pathological Hand Tremor Extraction

S Shahtalebi, A Mohammadi, SF Atashzar, RV Patel

2017 IEEE Global Conference on Signal and Information Processing Citations: 3

Talks & Contributions

Founder and Organizer of the "Ethical AI" Reading Group

An initiative to foster discussions on the ethical aspects of artificial intelligence.

PhD Representative at Council Meetings, Concordia University Canada

Participated in departmental council meetings, contributing to discussions and decision-making (2018-2020).

Reviewer and Contributor at Notable Conferences

Acted as a reviewer and contributor to discussions in various prestigious conferences such as MAIS, ICASSP, EMBC, SMC, and GlobalSIP.

Awards & Recognition

Postdoctoral Fellowship

Vector Institute, Toronto, ON, Canada (2021)

$75,000/year

Relève Étoile Louis-Berlinguet (Rising Star)

Fonds de Recherche du Québec (FRQ), Quebec, Canada (2020)

Concordia Stand-Out Graduate Research Award

Concordia University, Canada (2020)

Concordia Accelerator Award

Concordia University, Canada (2019)

$5,000

Concordia Merit Scholarship

Concordia University, Canada (2018)

$10,000

Concordia Graduate Student Mobility Award

Concordia University, Canada (2018)

$3,000

Tuition Award of Excellence for International Students

Concordia University, Canada (2016)

$36,000

Ranked 6th (top 1%) in Nationwide Exam for PhD

National Organization of Educational Testing, Iran (2016)

Rank 1st in M.Sc. program of Biomedical Engineering

K.N.Toosi University of Technology, Tehran, Iran (2015)

Distinguished paper in International Mathematics Competition

Tournament of Towns, Moscow, Russia (2008)

Contact

If you have any questions or would like to discuss potential collaborations, feel free to reach out to me.

(+1) 514-662-6571