Puja Saha

Puja Saha

Data Scientist at mlHealth360

About

I am a Healthcare AI and Data Scientist specializing in medical imaging, with experience building clinically deployable deep learning systems for radiology and hospital workflows. At mlHealth360, I develop advanced pipelines for automated chest X‑ray interpretation and report generation, integrating large‑scale real-world datasets with high-performance models optimized for both cloud and edge deployment.

I hold an M.A.Sc. in Computer Engineering (CGPA: 3.93/4.00) from the University of Guelph, supported by multiple competitive scholarships including the Queen Elizabeth II Graduate Scholarship in Science and Technology. My graduate research focused on Adaptive Differentially Private Federated Learning (ADP‑FL) for 3D medical segmentation, achieving up to 18% Dice improvement at extremely strict privacy budgets (ε = 0.001) across heterogeneous multi‑institutional datasets. This work earned the CAD Best Paper Award (1st Place) and Robert F. Wagner All‑Conference Best Paper (Finalist) Award at SPIE Medical Imaging-2026, Vancouver, Canada.

My broader interests span AI for Healthcare, including medical image analysis, privacy‑preserving learning, multimodal clinical AI, and robust model deployment. I am passionate about designing intelligent, secure, and reliable systems that meaningfully support clinicians and improve patient outcomes.

Featured Projects

Semantic Segmentation Project

Multiclass Semantic Segmentation

KiTS19 dataset segmentation with 91.03% kidney and 62.82% tumor Dice scores.

PyTorch MONAI OpenCV
View on GitHub →
Medical Chatbot Project

Medical Chatbot with RAG

PubMed-trained LLM with 96.7% context precision and 85% faithfulness.

LangChain Pinecone RAG
View on GitHub →
Job Market Analysis Project

Job Market Analysis & Prediction

Big data analysis of 1.6M job records with $10K RMSE salary prediction.

PySpark ML Data Science
View on GitHub →

Publications

Adaptive Differential Privacy for Federated Medical Segmentation Across Modalities and Complexity Levels CAD Best Paper: 1st Place R.F.W. All-Conference Best Paper: Finalist 2026

SPIE Medical Imaging, Vancouver, BC, Canada SPIE Proceedings

Convolutional Neural Network to Classify Medical Images of Rare Brain Disorders 5th Best Paper 2022

IEEE International Conference on Healthcare Engineering (ICHE), Johor, Malaysia IEEE Xplore

Photosweep: An Engineering Approach to Develop Cost Effective Sterilization System for Hospitals 2019

International Conference on Emerging Trends in Engineering (ICETE), Hyderabad, India Springer

Development of an Inexpensive Proficient Smart Walking Stick for Visually Impaired People 2019

International Conference on Emerging Trends in Engineering (ICETE), Hyderabad, India Springer

CV & Resume

Academic CV

Download CV

Professional Resume

Download Resume

Contact