Databricks announced the public preview of Mosaic AI  Agent Framework & Agent Evaluation alongside our Generative AI Cookbook at the Data + AI Summit 2024. 

These tools are designed to help developers build and deploy high-quality Agentic and Retrieval Augmented Generation (RAG) applications within the Databricks Data Intelligence Platform.  

Challenges with building high-quality Generative AI applications 

While building a proof of concept for your GenAI application is relatively straightforward, delivering a high-quality application has proven to be challenging for a large number of customers. To meet the standard of quality required for customer-facing applications, AI output must be accurate, safe, and governed. To reach this level of quality, developers struggle to 

• Choose the right metrics to evaluate the quality of the application
• Efficiently collect human feedback to measure the quality of the application
• Identify the root cause of quality problems
• Rapidly iterate to improve the quality of the application before deploying to production

Continue to read more.