Rikai: a new data format for analytics on unstructured data at scale

Data infrastructure for AI is broken. Repeated conversions, disjoint workflows, and manual QA of labels are just some of the pain-points experienced by any AI team trying to bring models to production. To solve these challenges, we've created a new open-source data format called Rikai. Rikai is designed specifically for AI teams to avoid the need to convert data between ETL, analysis, and training phases of the AI workflow. Using semantic-typing, Rikai also supports deep understanding of AI datasets using plain vanilla SQL. Finally, Rikai already works with existing analytics systems so it fits right into the modern data stack without the need for new compute frameworks or engines.