This repository serves as a comprehensive database of 450 machine learning system design case studies from over 100 companies, offering practical insights into how leading organizations like Netflix, Airbnb, and DoorDash implement ML. It's an invaluable resource for ML engineers, system designers, and researchers seeking to understand real-world ML applications and best practices.

How It Works

The repository compiles publicly available case studies, primarily sourced from company engineering blogs and technical publications. Each entry details a specific ML use case, including the problem addressed, the industry, a brief description, the title of the case study, relevant tags (e.g., "recommender system," "LLM," "fraud detection"), the year of publication, and a direct link to the original source. This structured format allows for easy browsing and analysis of trends in ML system design.

Quick Start & Requirements

No installation or specific software is required. The repository is a curated collection of links to external articles. Users can access the information directly through the provided table and links.

Highlighted Details

• Breadth of Applications: Covers a wide array of ML domains including recommender systems, LLMs, computer vision, NLP, fraud detection, demand forecasting, and operational efficiency.
• Industry Diversity: Features case studies from tech, e-commerce, media, finance, delivery, and gaming sectors.
• Recency: Includes numerous case studies from 2023 and 2024, reflecting the latest trends in ML, particularly the impact of Large Language Models (LLMs).
• Source Credibility: Links to reputable sources like Netflix Tech Blog, Uber Engineering, Google AI Blog, and Meta AI.

Maintenance & Community

This repository is a static compilation, with the last update indicated by the publication dates of the linked articles. The primary contribution is attributed to Evidently AI, with the repository acting as a curated mirror. There are no active community channels or roadmap for this specific repository.

Licensing & Compatibility

The repository itself does not host content but links to external articles. The licensing of the original content would be governed by the respective publishing companies. Compatibility for commercial use depends on the terms of service of each linked external article.

Limitations & Caveats

The repository is a curated list of links and does not provide code or implementation details. The depth of technical information varies significantly based on the original source article. The accuracy and completeness of the information are dependent on the original publications.