AI-Powered Recommendation Engine for a Leading Health App

The client is a popular weight loss app that focuses on changing behavior to adopt a healthy lifestyle and appearance. The app makes it entertaining to build new habits as part of a supportive community. Among other features, it enables users to listen to live sessions from wellness coaches, other experts, and leading voices.

To increase user engagement and loyalty, the client imagined building a recommendation engine as a core function of the app. Its key task would be an effective matching system between user goals and preferences AND relevant live streams and health-related trending topics.

Our team has developed an AI-powered recommendation engine that works with both audio and text formats. By leveraging a siamese-architecture-inspired network, the system considers multiple parameters to produce relevant and engaging content recommendations.

Audio content, with its multifaceted attributes, poses a significant challenge for recommendation systems. The answer itself, along with the question's topic, relevance, provided value, and wisdom, all play crucial roles in the process of suggesting.

All the inputs are usually recorded via phone. Thus, many things might contribute to overall audio quality — mic sensitivity, environmental noise, speech loudness, speed, pitch, etc.

A key requirement is the real-time generation of recommendations influenced by the user's current activity. The system must ensure that new content is instantly available. This requirement eliminates the possibility of caching or pre-computed recommendations. The system should be designed to handle a growing user base. Services should be scalable and able to adjust quickly to changes in user volume.

All recommendation systems have the cold start problem. It is usually related to new users who don't have historical activity. Such users should still be able to get hints right after signup and eventually should get a more personalized experience.

We trained an ML model to create embeddings for answers and users. The model features include audio transcription, question text, author description, and historical stats like listen rate, likes, comments, etc. Every day, the model is automatically retrained to tune weights and adjust embeddings based on the latest user's activity.

All answers on the platform were moderated by humans and scored according to rules. Those scores reflected how good the answers were, how precisely they answered the topic, and how much value they gave to the listener.

At the later stage of the project, we implemented an auto-scoring mechanism to reduce the amount of human work on the platform. We developed a comprehensive set of OpenAI ChatGPT prompts to evaluate the content score by the same criteria as humans did.

Additionally, we developed an audio processing pipeline to analyze the audio quality based on a set of features like speech ratio, short-time objective intelligibility (STOI), perceptual evaluation of speech quality (PESQ), signal-to-noise ratio (SNR), etc.

The system is primarily built using Python tech stack and scalable databases and is deployed on AWS Cloud. All the services are running on Lambdas and ECS. External integration with the app is performed via Kafka. Utilizing the OpenSearch and Milvus databases allows the creation of on-the-fly recommendations of 20-50 items in ~100-300ms time.

The recommendation engine serves as a key to the app's functionality ensuring the increase in daily active usage. Customers can easily find relevant content, be it audio clips, podcasts, chats, or trending topics.

Thanks to a truly curated experience, millions of users enjoy better content that impacts their lives positively. Due to robust architecture and infrastructure design, the app can sustain supporting a large user base without latency issues.

The consistent performance, great usability, and content relevancy ensure that the company consistently finds itself among the top players in digital health and weight management app ratings.