Earlier in the course, you saw a few ways to create a custom ML model. We'll briefly review them here, and point you to additional resources where you can practice building more yourself. One of the easiest ways to create a custom model on structured data from scratch, is to try out BigQuery ML. Earlier in this course, you practiced creating a machine learning data set and identifying features and labels. Choosing the right model type for your data set and what you are trying to predict or infer. Providing any custom model options, training the model and evaluating its performance, inspecting what the model learned about the weight of each feature and predicting an unknown future data. I'll provide additional resources and links for you to practice and learn more about BigQuery ML. Lastly, we mentioned that ML engineers often create their own models using open-source libraries, like TensorFlow running on GCP. The value of these models can be huge, if you build and train them correctly or minimal if they are not done well. If you're looking for experience building TensorFlow models on GCP, check out our ML on GCP specialization in our additional resources section of the course.