Sumit Pandey, PhD

Machine Learning Engineer | Data Scientist

Building production-ready AI systems that drive business value. Innovation-driven Data Scientist specializing in end-to-end ML solutions for computer vision and predictive analytics.

LinkedIn GitHub Medium
Sumit Pandey

🚀 Core Expertise

Machine Learning & GenAI

Deep Learning • Computer Vision • Data Analytics • NLP (LLM and agents)

MLOps & Engineering

Root Cause Analysis • Optimization • CNNs • Transformers • LLM fine-tuning (GPT-4.x, Hugging Face, ETL pipeline)

Data Domains

Medical Imaging • Predictive Maintenance • Business Analytics • Production ML Systems

Tools & Skills

Python • PyTorch • TensorFlow • Docker • Git • CI/CD (GitHub Actions) • Reproducible Pipelines (MLflow) • Linux/Bash • HPC cloud (Azure) • pandas • numpy • HuggingFace

💼 Professional Experience

Founder & Writer

Towards Deep Learning | ~18k readers | 120+ permanent members

Aug 2023 - Present
  • Translate complex Machine Learning and Data Science concepts and research into clear, practitioner-friendly insights/tutorials for a global audience
  • Built and managed a global community of ML practitioners and researchers

Transferable Skills: Mentoring • Simplifying complex topics • Storytelling • Audience engagement • Knowledge translation

External Consultant

University of Copenhagen, Denmark

Jun 2025 - Aug 2025
  • Co-developed Zero-MED-YOLO, a no-code solution that cut data preprocessing and model training preparation time from 3–4 hours to just 2–3 minutes by fully automating complex pipelines

Transferable Skills: Cross-functional collaboration • Machine learning • Software development • Technical consulting • Knowledge transfer

Visiting Researcher

Martinos Center, Harvard and MIT, Boston

Aug 2024 - Jan 2025
  • Developed an automated preprocessing pipeline for images and tabular data including filtering, stripping, registration etc.
  • Developed and implemented the MED-YOLO project, focused on advanced 2D and 3D image segmentation for sparse datasets

Transferable Skills: Data preprocessing • Process automation • Scalable pipeline dev. • Data-efficient modeling • Cross-functional collaboration

PhD Fellow

University of Copenhagen

Apr 2022 - Mar 2025
  • Led the Deep Consciousness project (R&D with Rigshospitalet, Copenhagen), developing predictive models for comatose patient outcomes using time-series imaging and tabular data, improving survival prediction accuracy from 0.75 to 0.82
  • Established a collaboration with Chang Gung Memorial Hospital (Taiwan) to develop a robust, automated point-of-care ultrasound aorta segmentation tool
  • Collaborated with doctors to refine deep learning methodologies, pinpointing critical gaps in the data preprocessing pipelines and implementing innovative analytical solutions to improve model outcomes
  • Applied classical statistical methods in Python, including model design, comparison, validation, and both regression and survival analysis for coma and brain tumor patient studies
  • Managed large hospital datasets on cloud-based HPC and Linux systems, using shell scripting (Bash) to train models

Transferable Skills: End-to-end ML development • Predictive modeling • Data preprocessing & automation • Statistical analysis & validation • Scalable data infrastructure • Cross-functional collaboration • Computer Vision

Research Assistant

Chang Gung Memorial Hospital, Linkou, Taiwan

Jul 2020 - Dec 2021
  • Performed cancer segmentation on MRI using state-of-the-art deep learning (CNN, GAN, U-Net), resulting in a published journal article in European Radiology (Impact Factor: 4.7)
  • Led the design and development of a Deep Learning course for hospital residents, bridging AI and medical practice

Transferable Skills: Deep learning • Computer vision • Curriculum design & training • Knowledge translation • Applied AI

🚀 Key Projects & Impact

Deep Consciousness

Developed predictive models for comatose patient outcomes using time-series CT imaging

Impact 0.82 AUROC, deployed at Rigshospitalet Copenhagen
Tech PyTorch, CNN, Survival Analysis, Time Series Modeling

Zero-MED-YOLO

No-code solution for medical image analysis with automated preprocessing

Impact 95% reduction in setup time (3-4 hours → 2-3 minutes)
Tech YOLO, Computer Vision, Process Automation

GPT-4 Customer Analysis

Automated sentiment analysis and classification system for customer feedback

Impact Reduced manual work from 30 hours to 1 hour per analysis cycle
Tech GPT-4.1, Gemma-2B, Python, GDPR-compliant deployment

Predictive Maintenance System

Optimized hemodialysis machine maintenance using genetic algorithms

Impact 60% cost reduction (~$34k per machine), maintained 99%+ availability
Tech Root Cause Analysis, Genetic Algorithms, Weibull Analysis

📚 Recent Publications

Integrating symbolic regression with physics-informed neural networks for simulating nonlinear wave dynamics in arterial blood flow

S Changdar, B Bhaumik, N Sadhukhan, S Pandey, S Mukhopadhyay, et al.

Physics of Fluids 36 (12) • 2024

Machine Learning Computer Vision

Validating YOLOv8 and SAM Foundation Models for Robust Point-of-Care Ultrasound Aorta Segmentation

S Pandey, CW Lu, CM Tan, PH Tsui, EB Dam, KF Chen

2024

Computer Vision Deep Learning

Comprehensive Multimodal Segmentation in Medical Imaging: Combining YOLOv8 with SAM and HQ-SAM Models

S Pandey, KF Chen, EB Dam

Proceedings of the IEEE/CVF International Conference on Computer Vision • 2023

Computer Vision Deep Learning

Fully automated segmentation and radiomics feature extraction of hypopharyngeal cancer on MRI using deep learning

YC Lin, G Lin, S Pandey, CH Yeh, JJ Wang, CY Lin, TY Ho, SF Ko, SH Ng

European Radiology 33 (9), 6548-6556 • 2023

Deep Learning Machine Learning

Advancing Kidney, Kidney Tumor, Cyst Segmentation: A Multi-Planner U-Net Approach for the KiTS23 Challenge

S Pandey, Toshali, M Perslev, EB Dam

International Challenge on Kidney and Kidney Tumor Segmentation, 143-148 • 2023

Computer Vision Deep Learning

Kvasir-Instruments and Polyp Segmentation Using UNet

KA Pandey Sumit

University of Oslo • 2021

Computer Vision Deep Learning
See More Publications on Google Scholar

🎓 Education & Awards

Education

PhD in Computer Science

2022-2025

University of Copenhagen, Denmark

Thesis: "From Deep Learning to Deep Consciousness"

MS in Electronic Engineering

2018-2020

Chang Gung University, Taiwan

CGPA: 3.6/4.0 | Published in European Radiology (IF: 4.7)

Awards & Recognition

Corti Hackathon Winner (2025)

Developed "Nurse Agent" for patient voice prioritization

PhD Fellowship Recipient

University of Copenhagen (2022-2025)

170+ Academic Citations

Established reputation in computer vision research

Published Author

4+ peer-reviewed journal articles in top-tier venues

🌐 Let's Connect

I'm always interested in discussing AI applications in healthcare, MLOps best practices, and opportunities to collaborate on impactful projects.

Email

sumitpandey171@gmail.com

Location

Copenhagen, Denmark

Available for

Consulting • Collaboration

Contact Me