Deepak Pathak

My work sits at the boundary between ML research and production engineering. I build systems that need to hold up scientifically—and actually run in the field.

At DFKI, I work on three main threads: multimodal crop-yield prediction from satellite and geospatial data, privacy auditing frameworks for ML models and LLMs under EU AI Act constraints, and the MLOps infrastructure (Kedro, Airflow, MLflow, CI/CD) that makes these systems reproducible and deployable with real partners.

Before research, I spent four years at IBM building distributed backend systems for telecom. That engineering foundation—reliable pipelines, clear interfaces, respect for scale—still shapes how I approach every project.

Full background →