01
July 2025 – Present
Project Engineer
- Built data drift detection across tabular, image, and text data: 8 supervised methods, 6 pre-trained unsupervised detectors for feature drift, and bounding-box count and area metrics for label drift.
- Developed data quality analysers for tabular, vision, and text data — 15 integrity checks for tabular data, outlier detection over 8 image properties, 7 feature-based metrics for text, plus feature-feature and feature-label correlation analysis.
- Ran mixed- and fixed-precision quantisation experiments for sharing model parameters across a federated learning setup built on Flower.
