Background: We Need Intelligent Big Data Stores. Traditionally, data stores have been managed primarily by database administrators with deep system expertise. We anticipate significantly broader demand from non-DB-expert users in the near future. For example, contemporary AI developers must routinely manage large and heterogeneous datasets such as text logs, graphs, vectors and images across iterative training and evaluation cycles. These workflows require laborious tasks including storage management, retrieval, and summarization tailored to various data types.
Siqiang Luo (Mon,) studied this question.