Parris is an open-source tool that automates the infrastructure setup for training machine learning algorithms on AWS. Published on GitHub with about 315 stars and updated in February 2026, it removes the repetitive server babysitting that surrounds ML training jobs.
Anyone who trains models knows the pattern: provision a server, configure it, start the job, then keep logging in over SSH to check progress. Parris is built to automate that loop. You bring an AWS account with credentials configured on your workstation, a machine learning algorithm to train, a dataset, and ideally an S3 bucket or similar storage for the results. Parris then handles launching the training stack following its getting-started and configuration guides, so the infrastructure side of an experiment becomes a repeatable procedure instead of a manual ritual. The project is written in Python and welcomes community contributions through its GitHub issues page.
The workflow is deliberately workstation-driven: you configure the tool locally, point it at the algorithm and the data for an experiment, and let it stand up the remote stack for you. For individual researchers, that removes the least enjoyable part of cloud training, remembering the exact sequence of manual steps that worked last time, and replaces it with configuration files that live next to the project and can be reused for every run.
Parris is free open-source software, released under the Apache 2.0 license according to the repository. You pay only for the AWS resources your training jobs consume.
Parris is for machine learning practitioners and small research teams who train on AWS and want scripted, repeatable training infrastructure without adopting a heavyweight MLOps platform.