In our experience, every nontrivial machine learning project is eventually stitched together by a set of bug-ridden and unmaintainable internal tools. These tools -- often a patchwork of Jupyter Notebooks and Flask apps -- are difficult to deploy, require reasoning about client-server architecture, and don’t integrate well with machine learning constructs like Tensorflow GPU sessions.

Streamlit is the first app framework built specifically for machine learning engineers. The open-source library enables you to quickly turn pure Python scripts into bespoke ML apps without any "app-building knowledge”. Under the hood, Streamlit statically analyzes Python code and generates computational dependency graphs, turning scripts into performant, interactive apps with minor annotation. Streamlit also hides the complexity of maintaining GPU sessions, websocket connections, client state, and threading for concurrent users. Growing only through word of mouth, Streamlit is already in use at several top machine learning institutions including Carnegie Mellon, University of Michigan, Uber, Stripe, Stitch Fix, The Allen Institute for Artificial Intelligence, and Google X. Our demonstration will lead the audience through building a tiny Streamlit app from scratch that analyzes self-driving car data, including running interactive object detection using the YOLO detector.