Machine Learning Operations (MLOps)

MLOps streamlines and automates the end-to-end machine learning lifecycle, covering model development, deployment, monitoring, and maintenance to ensure efficient and reliable ML systems.

Courses

• Machine Learning Engineering for Production (MLOps) Specialization (opens in a new tab) by DeepLearning.AI, equips learners with production engineering skills, bridging ML and software engineering. Key aspects include end-to-end ML production system design, tool proficiency, real-world problem-solving, and an applied learning project for practical experience. This specialization comprises four courses: Introduction to ML in Production, ML Data Lifecycle in Production, ML Modeling Pipelines in Production and Deploying ML Models in Production.

• Hugging Face on Amazon SageMaker (opens in a new tab) offers guidance on utilizing Amazon SageMaker for ML model training and deployment. It includes acomprehensive ecosystem for exploring research, search options for specific tasks, a simple launcher for popular LLMs, and resources for deploying models on SageMaker.

• Practical Data Science on the AWS Cloud Specialization (opens in a new tab) by AWS & DeepLearning.AI, empowers data professionals to build, train, and deploy scalable ML pipelines in AWS. It blends data science and cloud computing, covering end-to-end pipeline development, hands-on labs, Python and SQL proficiency, and a 10-week duration. Perfect for data-focused developers and analysts seeking efficient ML deployment on AWS. This specialization comprises three courses: Analyze Datasets and Train ML Models using AutoML, Build, Train, and Deploy ML Pipelines using BERT and Optimize ML Models and Deploy Human-in-the-Loop Pipelines.

While the above courses may not be accessible for free, you can seek financial aid to enroll in them.

Guides

• Cloud GPUs (opens in a new tab) by The Full Stack, In neural network training, hardware acceleration—typically GPUs—is crucial. The platform organizes cloud GPU vendor pricing into sortable tables, categorizing offerings into GPU Cloud Servers, which are long-running and possibly pre-emptible, and Serverless GPUs that scale to zero when inactive, similar to AWS Lambda or Google Cloud Function.

• MLOps Org (opens in a new tab) is a comprehensive MLOps resource, covering core principles, recommended books, frameworks, tools, and best practices. It discusses fundamental MLOps principles, references key frameworks and communities, and highlights crucial components for MLOps success. Additionally, it introduces the MLOps Stack Canvas for structured ML application development and maintenance.

Books

• Introducing MLOps: How to Scale Machine Learning in the Enterprise (opens in a new tab) by Mark Treveil & Nicolas Omont provides a comprehensive introduction to MLOps, explaining its key concepts, the problems it solves, and the processes involved in a successful ML life cycle.

• Practical MLOps: Operationalizing Machine Learning Models (opens in a new tab) by Alfredo Deza & Noah Gift, offers a hands-on approach to MLOps, focusing on the operationalization of machine learning models. It provides practical examples and techniques for managing the production life cycle of ML models using MLOps.

• Machine Learning Engineering (opens in a new tab) by Andriy Burkov's book, while not exclusively centered on MLOps, delves into diverse facets of ML engineering, encompassing model deployment, monitoring, and maintenance. It serves as a comprehensive resource for grasping the engineering elements of ML, making it a valuable complement to your MLOps expertise.

• Designing Machine Learning Systems: An Iterative Process for Production-Ready Applications (opens in a new tab) by Chip Huyen, emphasizes crucial design choices in creating and implementing machine learning systems, encompassing MLOps tools and infrastructure. It offers a comprehensive approach to design, covering topics from data and feature engineering to model development, deployment, testing, and MLOps, ensuring reliability, scalability, and adaptability.

Reference

• Google Cloud Developer Cheat Sheet (opens in a new tab) is a quick reference guide for developers on Google Cloud, featuring every product, feature, and service in the Google Cloud family in depth. Google Cloud's suite includes compute, storage, networking, big data, AI/ML, IoT, management, security, and developer tools. (poster on GitHub)

• Weights & Biases Docs (opens in a new tab): Weights & Biases is the machine learning platform that enables developers to accelerate model development. With just a few lines of code, W&B lets you track, compare, and visualize your ML models. This simplifies experiment tracking for better analysis and result reproducibility.