Tradition ML development methodology does not enable a large number of contributors, each with distinct objectives, to work collectively on the creation and extension of a shared intelligent system. Enabling such a collaborative methodology can accelerate the rate of innovation, increase ML technologies accessibility and enable the emergence of novel capabilities. We believe that this can be achieved through the definition of abstraction boundaries and a modularized representation of ML models and methods. We present a multi-agent framework for collaborative and asynchronous extension of dynamic large-scale multitask intelligent systems.