Research Areas: Deep learning for applied ML, Interpretability, Machine learning theory

- Leveraged entropy information between ad session attributes to detect surprising attribute value patterns suggestive of invalid traffic (IVT)- Analyzed behavioral signals from ad sessions to detect bots and other IVT modes using specialized classifiers

- Optimized and maintained a large-scale image classification pipeline for multiclass tasks on specially-structured images- Applied deep learning, curated a feature extraction pipeline, and incorporated Bayesian hyperparameter optimization to increase classifier performance