How Peloton Uses AWS to Power Personalized Fitness for 6.5 Million Members

Peloton, a leader in connected fitness, delivers personalized workout recommendations to its 6.5 million members across eight platforms. This is achieved through a sophisticated architecture built on Amazon Web Services (AWS).

How Peloton Uses AWS to Power Personalized Fitness for 6.5 Million Members

Peloton has revolutionized the fitness industry by providing personalized workout experiences to millions of users. At the heart of this experience lies a robust and scalable architecture built on Amazon Web Services (AWS). This system enables Peloton to deliver tailored workout recommendations to its 6.5 million members across a wide range of platforms.

The Architecture:

Peloton’s architecture on AWS is designed for efficiency and personalization. Here’s a breakdown of the key components:

• Model Training: Peloton trains its recommendation models daily, using historical member data stored on Amazon S3. Amazon EMR clusters play a crucial role in preprocessing this data and training the models. The trained models are then stored in a model registry. Job orchestration is managed using Airflow on Amazon Managed Workflows for Apache Airflow (MWAA).
• Data Storage: To ensure low-latency data retrieval, Peloton employs a graph database and Amazon DynamoDB with a feature store. The graph database stores information about members and their class history, while the feature store contains pre-processed data, including member preferences and historical behaviors.
• Recommendation Process: When a user initiates a workout, a hydration service retrieves recommendations. This service calls a recommendation service, which orchestrates the entire process. The recommendation service fetches features from the feature store, candidate classes from the graph store, and the latest model. An inference service then scores the candidates, and the personalized recommendations are delivered to the user.
• Feedback Loop: The recommendation service logs data for training purposes, creating a feedback loop that continuously improves the system. This valuable data is stored in Amazon S3, enabling Peloton to refine its models and enhance the user experience.

By leveraging the power and scalability of AWS, Peloton has created a sophisticated system that delivers personalized workout recommendations to millions of users. This architecture allows Peloton to adapt to individual preferences and provide a truly tailored fitness experience.