Work on methods to reduce the greenhouse gas emissions of deep learning models trained in a federated fashion under non-IID assumptions, correlated and heterogeneous participation, under the supervision of Prof. Dr. Giovanni Neglia.
Tweaked and expanded the transformer-based wildfire detection model trained on satellite images (VIIRS) to encompass new geographic regions.
Critically evaluated VAE-based deep learning models for anomaly detection in time series, emphasizing the need for better benchmarks and a stable metric over the "point-adjusted" standard
Assisted in tutorials, graded assignments, and prepared course materials for Advanced Machine Learning, Artificial Intelligence, and Search Engines courses
Implemented machine learning solutions for pharmaceutical manufacturing; Developed fault detection systems using statistical and ML approaches
Scaled recommender engine to process 7M+ customer transactions using Apache Spark; Utilized a combination of n-gram and Cologne phonetics algorithms for efficient entity resolution
Python, Java, Scala, TensorFlow, PyTorch, Keras, Apache Spark, Airflow, OpenMPI, SQL, AWS, GCP, Git, Slurm, OAR