161: The Intersection of Generative AI and Data Infrastructure with Chang She of LanceDB
The Data Stack Show
00:00
Separation of Transactional Store and AI Feature Store
Having a production transactional store also function as the AI feature store and serving store has been seen in some installations, especially in small scale operations. However, as companies scale up, it becomes less effective due to differing workload requirements. AI serving stores supporting vector search workloads often have stringent requirements and are more CPU intensive. Mixing the transactional and AI feature store creates challenges for both workloads. At scale, many companies find it beneficial to separate the two stores for improved efficiency.
Transcript
Play full episode
