The roadmap page lists topics both scheduled to be implemented and/or explored. Some of the topics are ongoing.

• Add an option to download autodoc.

• Make Steam utility functions and an option to specify the instance in runtime.
• Train, test and validate the model using Pandas.DataFrame.

• Incorporate a validation dataset option for the model building process.
• Support choosing of columns for categories, features and dropping for the model building process.

• Provide hyper-parameters for the model building.

• Integrate H2O DriverlessAI client with current codebase. Unify the API and make DAI usable locally if available.
• Support H2O Steam, if available, to instantiate DAI. We should be able to spin both plane and multinode instances.
• Integrate OpenID with WaveML so DAI can be used both with user/password and token approach.
• Provide a way to deploy a model and predict on the deployed model.
• Move API from GitHub frontpage to a more appropriate place.

• Explore the possibilities to spin H2O-3 on H2O Steam.
• Explore the model deployment process using MLOps.
• Look at the MOJO and Python scoring pipeline as the alternative way of loading a model into Wave ML.
• Provide a way to pass a DAI/H2O-3 client for the user.
• Explore the advantages of having a regular Wave component able to handle building and deploying a model.

• Explore the advantages of automatic data typing for the rest of the columns (other than a target column).
• Make other dataframes working with Wave ML - H2OFrame and datatable.
• Provide a way to explain a model. (Might be not possible from MLOps yet).
• Create a program able to generate a template Wave App based on a given dataset. Useful to quickly start a predictive app.

• Set up automatic testing to test Steam->DAI->MLOps integration.

• Add more examples to the repository. We need both decision-support and ML-oriented examples.
• Create a chart with decision processes being made inside WaveML. It should be clear what resource will be used in the background based on ENV variables.
• Create an architecture diagram also showing the OIDC flow.
• FAQ section in Wiki.