ML Engine - Machine Learning in the Cloud

TL;DR

• Learn to build and deploy custom ML models on GCP using Python, Firestore, ML Engine, and Firebase Cloud Functions.

Transcript

today we'll start from zero and finish the video with our own custom machine learning model deployed to the cloud as an API we will pull some actual user data that we have saved in the firestore database then we'll export it to data lab so we can train a Python based machine learning model with virtually unlimited compute resources once we have a s... Read More

Key Insights

• 😄 Python is an essential language for building machine learning models due to its ease of use and robust libraries like scikit-learn.
• 👤 Firestore serves as a valuable source of data for training predictive models that enhance user experience.
• 🌥️ Data Lab on GCP provides powerful computational resources for processing and training large datasets in the cloud.
• 🚂 Scikit-learn offers tools for data preprocessing, cleaning, and algorithm training essential for machine learning tasks.
• 🌍 ML Engine enables versioning and deployment of machine learning models as APIs for real-world applications on Google Cloud.
• 👻 Firebase Cloud Functions allow for the creation of serverless APIs to interact with deployed ML models.
• 🏆 Insomnia, a useful tool for testing API endpoints, validates the functionality and performance of deployed machine learning models.

Questions & Answers

Q: What are the key components involved in building a custom machine learning model on Google Cloud?

The process includes data collection from Firestore, training the model in Data Lab, deploying it with ML Engine, and creating an API with Firebase Cloud Functions.

Q: How is the data cleaned and prepared for training the machine learning model?

Scikit-learn is used to clean the data and convert string values to numerical classes, enabling the model to make predictions based on the features.

Q: Why is testing data separated from training data in machine learning?

Testing data is separated to validate the model's performance and ensure it can generalize to unseen data, preventing overfitting during training.

Q: How is the machine learning model deployed as an API for real-world use?

The model is saved as a Joblib file in Firebase Storage, then served through ML Engine, allowing predictions to be made in milliseconds via a cloud function.

Summary & Key Takeaways

• Using Python, Firestore, and GCP services, the video demonstrates building a custom ML model from scratch.

• The process includes data collection from Firestore, training the model in Data Lab, deploying it with ML Engine, and creating an API with Firebase Cloud Functions.

• Key steps involve cleaning data with scikit-learn, training a random forest algorithm, and deploying the model as an API using Google Cloud services.