A data scientist describes the practical workflow of building machine learning systems, emphasizing that modeling is only about 10% of the job. The majority involves data cleaning, exploratory data analysis (EDA), feature engineering, understanding business context, and productionizing models with tools like Docker, ML Flow, FastAPI, AWS, and Evidently. The use case highlights real-world challenges such as handling messy and semi-structured data, choosing appropriate evaluation metrics for imbalanced datasets (e.g., fraud detection), and the need for production-grade coding and scalable deployment.

Jun 7, 2026, 7:00 PM