Cloud Native Machine Learning

Getting a model to work in an experiment is one thing; building the cloud-based systems that prepare its data, train it, and serve its results is another. Cloud Native Machine Learning tackles that production gap through the practical lens of serverless machine learning, with Amazon Web Services as its main cloud example.

This Version 5 manuscript is organized around a structured-data use case. It takes readers from dataset preparation into model development and then toward the engineering practices that support a working machine-learning pipeline.

From experimental code to a production platform

The opening section explains why production machine learning involves much more than writing model code. Data storage, processing, infrastructure, and serving all have to work together. The manuscript introduces cloud services and a serverless approach as ways to assemble those supporting capabilities while focusing less on managing underlying infrastructure.

Prepare the data before the models

Part 1 centers on mastering the dataset: importing, exploring, and preparing structured data for machine learning. That emphasis gives the modeling work a practical foundation, connecting data quality and preparation with the decisions that follow.

Work with PyTorch in the cloud

In Part 2, the manuscript introduces PyTorch concepts including tensors, automatic differentiation, and optimizers, then addresses training at scale and distributed training. It also explores assessing alternative model implementations and deploying trained models to respond to user demand.

Build the pipeline around the model

The final part turns to feature engineering, hyperparameter tuning, and assembling machine-learning pipelines. Together, these subjects show how data preparation, model training, and deployment fit into a broader production workflow rather than standing as isolated tasks.

Who will find it relevant?

The manuscript is intended for readers with experience in Python and SQL data analysis and some familiarity with machine learning. It is especially relevant to practitioners looking to strengthen the engineering side of supervised learning with structured data. Appendixes introduce machine learning, Docker, and public cloud computing for readers who want context on those supporting topics.

For readers ready to think beyond the notebook, this manuscript offers a practical route into the cloud infrastructure and pipeline work that helps machine-learning projects reach production.