This repository provides an end-to-end MLOps example on Google Cloud, leveraging Vertex AI, TensorFlow, and TFX. It's designed for engineers and data scientists looking to implement a robust machine learning workflow, from data management and experimentation to deployment and monitoring, all within the Google Cloud ecosystem.

How It Works

The project demonstrates a comprehensive MLOps lifecycle using a Chicago Taxi Trips dataset for tip prediction. It utilizes Keras for model development, TFX for building and orchestrating training pipelines, and Vertex AI for managed training, hyperparameter tuning, model registry, endpoint deployment, and monitoring. Data processing is handled by Dataflow, and CI/CD for pipeline and model deployment is managed by Cloud Build, with Cloud Functions and Pub/Sub enabling continuous training triggers.

Quick Start & Requirements

• Install: Clone the repository and install dependencies using pip install tfx==1.2.0 --user and pip install -r requirements.txt.
• Prerequisites: Google Cloud account, AI Notebook instance, google-cloud-sdk (updated).
• Dataset: Chicago Taxi Trips dataset (available via BigQuery).
• Links: Official Documentation

Highlighted Details

• End-to-end MLOps workflow covering data management, experimentation, training formalization, operationalization, deployment, prediction serving, and monitoring.
• Integration with Vertex AI services including Datasets, Experiments, Metadata, TensorBoard, Pipelines, Model Registry, Endpoints, and Model Monitoring.
• Use of TFX for pipeline orchestration and Cloud Build/Cloud Functions/Pub/Sub for CI/CD and continuous training.

Maintenance & Community

This is sample code provided for educational purposes and is not an official Google product. It is licensed under the Apache License, Version 2.0.

Licensing & Compatibility

Licensed under the Apache License, Version 2.0. This license permits commercial use and linking with closed-source projects.

Limitations & Caveats

The README notes that pip dependencies issues can be ignored, implying potential version conflicts or outdated dependencies that may be resolved in later TFX versions. The project is explicitly stated as sample code for educational purposes.