Sanaz Kaviani
Ai Researcher @Jubilant Radiopharma
Signup · Get unlimited contacts
WORK HISTORY
Ai Researcher @Jubilant Radiopharma
Montreal, QC, CA
Developed and trained diffusion-based deep learning models (3D-DDPM and 3D-LDM) for denoising low-count cardiac Rubidium-82 PET images and synthesizing high-dose equivalents- Implemented models using OpenAI’s Guided Diffusion (3D-DDPM) and ControlNet with 3D-LDM (from Hugging Face)- Trained with text conditioning across six categories (rest/stress × 3 noise levels)- Achieved up to 20× faster inference using Rectified Flow and scaled training on Slurm-managed HPC clusters using Torch Distributed Data Parallel (DDP)- Libraries/Tools: PyTorch, Hugging Face Diffusers, Torch DDP, Slurm, TensorBoard Developed AI-based parametric mapping (kinetic modeling) and denoising pipelines for Rubidium-82 cardiac PET using a dataset of real patients- Built a self-supervised, physics-informed transformer network in TensorFlow to predict kinetic parameters from time-series PET data- Conducted evaluation using Cedars QBS/QGS, 4DM, and FlowQuant, and validated performance via TPD/iMFR, bootstrapping, permutation tests, and DeLong’s test- Libraries/Tools: TensorFlow, lstatkit, Invia’s 4DM, FlowQuant Developed U-Net Transformer models for automated segmentation and mismatch detection in ventilation/perfusion (V/Q) lung scans- Compared a custom 3D Bottleneck Transformer-UNet to SwinUNETR, DynUNet, and baseline UNet- evaluated via Dice coefficient and ROC analysis against expert segmentations- Libraries/Tools: MONAI, PyTorch, SwinUNETR, DynUNet, Scikit-learn Collaborated with cardiologists, physicists, and industry stakeholders to ensure the clinical relevance and utility of AI-driven imaging solutions. Contributed to data curation, versioning, and labeling pipelines, and integrated imaging-derived metrics into clinical research studies and trial data systems. Authored and contributed to technical documentation for product reliability and FDA regulatory clearance, supporting evidence generation and reproducibility in AI-based imaging products.
EDUCATION
Azarbaijan University
Bachelor of Applied Science - BASc, Physics
Shahid Beheshti University
Master’s Degree, Medical Radiation Engineering
Université de Montréal
Doctor of Philosophy - PhD, Biomedical/Medical Engineering
SKILLS
ABOUT SANAZ KAVIANI
Clinical Data Scientist and AI researcher with over 10 years of experience in medical imaging, deep learning, and clinical data analysis. Skilled in PET and MRI image reconstruction, denoising, segmentation, and kinetic modeling, with expertise in transformer-based and diffusion models for improving image quality and extracting quantitative biomarkers.Ph.D. in Biomedical Engineering from Université de Montréal, with work spanning large-scale clinical datasets, cloud/HPC computing, and regulatory documentation for FDA-aligned AI products. Published author, patent holder, and recipient of multiple research awards, with AI methods validated and applied in real-world clinical settings. Dedicated to bridging advanced AI research with practical healthcare applications.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.