Graphlet Analysis Insights
Doris discusses the challenges of analyzing complex data pipelines and introduces the concept of graphlets to encapsulate model training execution. Surprising findings reveal that over 30% of graphlets fail to push models to production, primarily due to data drift and workflow structure changes, rather than model type or code alterations. A machine learning model successfully predicts these failures, reducing wasted computation by over 50%.In this clip
From this podcast
Super Data Science: ML & AI Podcast with Jon Krohn
SDS 573: Automating ML Model Deployment — with Doris Xin
Related Questions