Data pipeline and access across multiple machine learned models

Abstract

The present disclosure describes systems and methods for storing incoming data and providing access to that data to multiple machine learned models in a data type-agnostic and programming language-agnostic manner. Operationally, a computing device may receive in coming data (e.g., from sensors, etc.). The computing device may store the incoming data in memory blocks, and index the memory blocks with a unique index (e.g., tag). The index may correspond to a determined tier for the memory blocks, and may enable the system to both locate the data once stored and enable the system to read (or use) the data upon receiving, for example, a data access request. In this way, systems and methods described herein provide for a robust data access and transfer mechanism that allows data to be stored a single time, but accessed by one or more different applications, machine learned models, and the like, simultaneously.