As the complexity of state-of-the-art deep learning models increases by the month, implementation, interpretation, and traceability become ever-more-burdensome challenges for AI practitioners around the world. Several AI frameworks have risen in an effort to stem this tide, but the steady advance of the field has begun to test the bounds of their flexibility, expressiveness, and ease of use. To address these concerns, we introduce a radically flexible high-level open source deep learning framework for both research and industry. We introduce FastEstimator.