Troubleshooting Unstructured Data with Embeddings

Explore the power of embeddings in troubleshooting unstructured data models in this insightful 32-minute conference talk from the Toronto Machine Learning Series (TMLS). Dive into the challenges faced by ML teams working with unstructured data, including images, text, and audio, which constitute 80% of generated data. Learn how internal embedding representations can provide valuable insights into deep learning models' inner workings. Join Amber Roberts, ML Engineer at Arize AI, and Kyle Gallatin, Senior Software Engineer I, Machine Learning at Etsy, as they share Etsy's journey with embeddings, discuss encountered challenges, and offer best practices for effectively troubleshooting unstructured data models. Discover how embeddings can be leveraged to identify issues, implement solutions, and continuously improve both models and data, ultimately enhancing the efficiency and effectiveness of ML workflows.

Troubleshooting Unstructured Data with Embeddings