Although there are several tools for building and deploying ML models, there are far fewer to help maintain models in production. Debugging production ML pipelines can be exceedingly challenging, so Shreya Shankar (an Amplify EIR) released mltrace. mltrace helps ML engineers trace a model’s output to its most upstream raw data file. It provides an API to log versions of data and pipeline components, a database that stores information on component runs, and a UI to visualize traces of steps.