TrueFoundry serves the entire lifecycle of a datascientist starting from running a Jupyter Notebook, training the models, logging the models to a model registry and then deploying the model in batch or realtime mode.

​
The ML Lifecycle

1

Start a Jupyter Notebook or Connect Local VSCode/Cursor to remote compute

To start a Jupyter Notebook, follow the docs here: Launch Jupyter Notebook.

To start a remote SSH server and develop on VSCode by connecting to remote compute, launch a SSH server

2

Run Your Training Job

Notebooks are great for experimentation and in some cases, can suffice for running small training jobs - however, in most cases, you will want to deploy the training job to a remote server where you can run it manually or on a schedule. Use the Jobs feature in TrueFoundry to create training job.

3

Log Your Trained Model and Metrics

You can log your models in the Truefounry model registry using the Python SDK or UI. You will need to first create a repository to store the models in. You can log your models using this guide.

4

Deploy your model

You can deploy the models either in realtime or batch mode. To deploy it in batch mode, use the Jobs feature. To deploy the models in realtime mode, use the Services feature in Truefoundry.