Deepak Pathak
🔬

Hi, I'm

Deepak Pathak

Machine Learning Engineer & AI Researcher at DFKI

Open to Senior ML Engineer, Applied Scientist, and Research Engineering roles. At DFKI, I build multimodal ML systems, earth observation pipelines, and privacy-aware production AI infrastructure backed by peer-reviewed publications across CVPR, remote sensing, and IEEE venues.

View Experience Contact
About

I work at the intersection of ML research and production engineering, building systems that need to be scientifically credible and operationally reliable.

At DFKI, my work spans three main threads: multimodal crop-yield prediction from satellite, weather, and geospatial data; privacy auditing for ML models and LLMs in regulated settings; and the MLOps infrastructure that makes these systems reproducible and deployable with real partners.

That work has led to R² up to 0.83 at field resolution, deployments with industry partners in Germany, and peer-reviewed publications spanning CVPR 2026, Remote Sensing of Environment, Computers and Electronics in Agriculture, and IEEE venues. Before DFKI, I spent four years at IBM building distributed telecom systems, and that engineering foundation still shapes how I approach every project.

Full background →

Experience

AI Researcher

DFKI — SmartCut

Lead DFKI-side execution for SmartCut, coordinating project partners while building Kedro pipelines for harvester-data harmonization and Earth Observation acquisition. Developed Apache Airflow DAGs for raw-data monitoring and automated pipeline execution for field operations.

AI Researcher

DFKI — MissionKI

Architected privacy auditing workflows for classification models and LLMs using membership inference and model inversion evaluations. Automated MLflow-based testing and reporting for healthcare AI compliance and supported domain-specific LLM fine-tuning and privacy evaluation.

AI Researcher

DFKI — Yield Consortium

Engineered multimodal crop-yield prediction systems across satellite, weather, soil, and geospatial features, reaching R² up to 0.83 at field resolution. Built scalable preprocessing, Dash dashboards, and MLflow/GitLab CI/CD/Docker workflows used with industry partners in Germany.

Machine Learning Research Intern

Miele & Cie. KG

Built deep metric, contrastive, and self-supervised vision models for fine-grained classification, using PyTorch Lightning and Azure Databricks for scalable training.

Student Research Assistant

virtUOS, University of Osnabrück

Contributed to the SIDDATA digital assistant by supporting backend infrastructure and neural recommender integrations.

Working Student — Software Developer

Aitech Concept UG

Implemented TensorFlow object detection models and Django applications for real-time tracking from surveillance feeds.

Application Developer

IBM

Developed distributed Java EE services in a telecom SOA environment and built Oracle BPM workflows deployed on Oracle WebLogic Server. Also implemented Spring Boot and Twilio automation for incident flagging to speed operational response.

Education

M.Sc., Cognitive Science

Osnabrück University

Specialization in machine learning, computer vision, and data science.

B.Tech, Electronics Engineering

HBTI Kanpur (UPTU)

Technical Skills

Frameworks, platforms, and methods I rely on most

Machine Learning

PyTorch
expert
Hugging Face
advanced
TensorFlow
intermediate
Multimodal Learning
expert
LLM Fine-tuning
advanced

MLOps & Pipelines

Python
expert
Kedro
expert
MLflow
expert
Apache Airflow
advanced
Docker
advanced

Cloud & DevOps

GitLab CI/CD
advanced
AWS SageMaker
advanced
Azure Databricks
intermediate
Linux
advanced
SQL
advanced

Remote Sensing & Trustworthy AI

Sentinel-2
expert
Earth Observation
expert
Computer Vision
advanced
Privacy Auditing
advanced
Selected Publications
M. Miranda, D. Pathak, P. Helber, B. Bischke, H. Najjar, F. Mena, C. Sanchez, A. Pai, D. Arenas, M. Valdenegro Toro, M. Charfuelan, M. Nuske, A. Dengel (2026). YieldSAT: A multimodal benchmark dataset for high-resolution crop yield prediction. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026.
Project DOI
F. Mena, D. Pathak, H. Najjar, C. Sanchez, P. Helber, B. Bischke, P. Habelitz, M. Miranda, J. Siddamsetty, M. Nuske, M. Charfuelan, D. Arenas, M. Vollmer, A. Dengel (2025). Adaptive fusion of multi-modal remote sensing data for optimal sub-field crop yield prediction. Remote Sensing of Environment, Vol. 318, 2025.
DOI
H. Najjar, D. Pathak, M. Nuske, A. Dengel (2025). Intrinsic explainability of multimodal learning for crop yield simulation. Computers and Electronics in Agriculture, Vol. 239, 2025.
DOI
M. Miranda, D. Pathak, M. Nuske, A. Dengel (2024). Multi-modal fusion methods with local neighborhood information for crop yield prediction at field and subfield levels. IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 2024.
DOI
C. Sanchez, D. Pathak, M. Miranda, M. Charfuelan, P. Helber, M. Nuske, B. Bischke, P. Habelitz, N. Rahman, F. Mena, H. Najjar, J. Siddamsetty, D. Arenas, M. Vollmer, A. Dengel (2023). Influence of data cleaning techniques on sub-field yield predictions. IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 2023.
DOI
D. Pathak, M. Miranda, F. Mena, C. Sanchez, P. Helber, B. Bischke, P. Habelitz, H. Najjar, J. Siddamsetty, D. Arenas, M. Vollmer, M. Charfuelan, M. Nuske, A. Dengel (2023). Predicting crop yield with machine learning: An extensive analysis of input modalities and models on a field and sub-field level. IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 2023.
DOI

Contact

Open to Senior ML Engineer, Applied Scientist, and Research Engineering roles, as well as collaborations and industry conversations.

Based in

DFKI GmbH

Kaiserslautern, Germany

View on Map

Reach me

Email

deepak_kumar.pathak@dfki.de

Find me on