Summary

In this chapter, we discussed quick ways to create a baseline model and demonstrated how that increases productivity.

We demonstrated MLflow functionality that supports MLOps and helps track model training and tuning. We also covered more complex classification frameworks that can be used in the lakehouse. Access to these frameworks made it possible to implement a DL model in PyTorch for the Parkinson’s FOG example. The openness of Databricks opens the doors for open source and proprietary innovations with API integrations, as shown by the SQL bot LLM. This integration saved time by not recreating the wheel and putting the SQL tool in the hands of our analysts sooner.

The next chapter will focus on moving our models into production.