At NatWest Group, I automated metadata extraction with Python, PySpark, and an LLM-driven DUO API pipeline, reducing manual documentation effort by 70%.
I also built a Spark lineage pipeline with Spline, OpenLineage, and Airflow, making transformations visible in NatWest’s Infomap UI and reducing dependency-tracing time by about 40%.
For the Finance and Risk Modelling team, I delivered privacy-safe data pipelines and developed a Spark-based data-masking utility adopted by 15–20 teams. I also built Cloudblitz, an internal tool for accessing AWS services from an IDE, and a CI/CD pipeline that cut deployment time by 75%.
At the University of California, San Diego’s Video Processing Lab, I developed an OCTA image-classification pipeline for wet AMD treatment outcomes that achieved 82% test accuracy. This work was published in Scientific Reports.

