Mohammad Joudy

Integrated, documented tools into Galaxy, an open-source, no-code platform that enables users to build and run reproducible ML workflows: • Integrated CleanLab for automatically improve data quality, achieving up to 28% increase in model accuracy on benchmarks by xgboost. • Adapted a deep learning (YOLOv8) object detection model to enable marine species analysis on image data by Earth scientists.

Developed a novel algorithm to help biologists estimate brain networks, supporting research to understand brain disorders: - Deployed to HPC an ML pipeline based on multivariate regression (VAR) to estimate neuronal networks, improved accuracy by 33%. - Designed a custom feature engineering method to enhance spike train detection, leading to a 40% improvement in model accuracy. - Tutored in a team, courses for more than 40 Masters students of Neuroscience and Biology: Quantitative Methods, Python.

Devised a generalized algorithm to parse PDFs with diverse layouts, storing content based on titles and subtitles. Acquired foundational NLP knowledge and programmed a RAG application by creating a retrieval model and training an LLM