Databricks has announced new synthetic data capabilities aimed at improving the evaluation of AI agents. This tool will enable developers to generate high-quality artificial datasets, streamlining the process of assessing the performance of AI systems.
The initiative is designed to reduce the time spent on consultations with subject matter experts, facilitating quicker deployment of AI agents in enterprises. Internal tests by Databricks suggest that this new offering can significantly enhance agent performance across various metrics.
Previously, Databricks integrated technology from its acquisition of MosaicML to strengthen its Data Intelligence platform. This integration supports the development of compound AI systems capable of reasoning, responding accurately, and performing tasks such as managing support tickets and making reservations.
The newly launched synthetic data generation API expands the existing Agent Evaluation capabilities, which already allow users to create defined datasets and assess AI agents' responses. This API aims to further refine the evaluation process, ensuring better performance of AI systems in diverse applications.