OpenAI faces significant challenges in data retrieval, a necessity for its 3,500 employees accessing 70,000 datasets that generate 600PBs of data daily. An internal query can be complex, leading to time-consuming discussions among analysts. Bonnie Xu from OpenAI’s data productivity team introduced Kepler, a tool that streamlines these data retrieval processes. Originally designed for data scientists, its user base expanded to finance and HR personnel. Kepler uses GPT-5 to seamlessly interact with various internal data sources via Slack and IDEs. It automates complex SQL queries, enabling quick responses to user inquiries, such as taxi travel time analytics or spikes in ChatGPT user activity. The tool also incorporates metadata for context, ensuring accuracy and continuous learning. Though Kepler won’t be open-sourced or offered as an enterprise solution, it dramatically increases productivity by simplifying data access and enhancing decision-making efficiency.
Source link
Share
Read more