Overview
Arré Voice wanted to enhance user engagement on its digital media platform with personalized content recommendations. The company needed a scalable solution capable of processing large volumes of user data while supporting artificial intelligence (AI)-driven algorithms for content personalization. By transitioning to a microservices architecture, using Amazon Web Services (AWS), and implementing Astra DB, a vector database from AWS Partner DataStax, Arré Voice enabled real-time data handling and improved operational efficiency. Astra DB’s query capabilities and developer tools facilitated the rapid development of its recommendation engine, which now generates over 30 million personalized suggestions per month with a 90-percent reduction in latency.
Opportunity | Enhancing Engagement through Personalization
As a user-generated content platform with an algorithmic feed, Arré Voice recognized the power of personalized recommendations. The India-based company wanted to enhance user engagement by transitioning its platform to a more advanced architecture. This would allow Arré Voice to implement AI-driven algorithms to better analyze user data and provide tailored content. However, Arré had complex needs, including accurate transcription and translation of audio data into a machine-readable format for data analysis. To achieve real-time personalized recommendations, the platform required a scalable and efficient database capable of managing substantial user data volumes and supporting quick queries, butdetermining the right solution for its goals was difficult. Further complicating personalization, Arré creators record content in English and Hindi along with four other Indian languages. Because it is a startup, these challenges were expected to grow as Arré Voice scaled operations.
Arré initially operated on a monolithic architecture, which made it difficult to scale and innovate quickly. As it transitioned into a microservices architecture, Arré Voice initially chose a cloud-based database that quickly presented limitations. The solution was costly and had a major learning curve. The team struggled with defining service boundaries and managing interdependencies and lacked the integration capabilities needed for AI-driven solutions. Arré needed to replace its new database—fast—with a scalable and efficient vector database capable of handling large volumes of data and supporting real-time queries. It was also essential that the platform seamlessly integrated with the existing infrastructure.
Solution | Building on a Microservices Architecture for Real-Time Insights
To modernize Arré’s architecture and address its complex data needs, Arré Voice developed 25 microservices, orchestrated using Amazon Elastic Container Service (ECS) for scalable deployment and simplified management. The company selected DataStax Astra DB as the core vector database, which offers real-time query execution and low-latency performance so it can process vast amounts of user data instantly. This enabled Arré to deliver personalized recommendations in real time, enhancing user engagement through context-sensitive content delivery. Astra DB’s cloud-native design ensured seamless integration into Arré Voice’s existing infrastructure, as well as easy scaling to handle increasing data loads without the operational complexities of traditional databases. Robust developer tools from Astra DB also empowered Arré Voice’s team to fine-tune recommendation algorithms, enabling ongoing improvements and rapid feature deployment. Astra DB’s ability to support complex queries and handle high volumes of diverse data was essential in creating precise and personalized recommendations for Arré users.
AWS infrastructure provided ongoing scalability and reliability. To handle complex workflows seamlessly and optimize precision, Arré used a combination of high-performance AWS services. For high-throughput data operations, the company integrated Amazon DynamoDBas a separate database solution to accelerate indexing and querying across the microservices. Amazon Neptune was introduced to map relationships in user data, further refining the precision of recommendations, while Amazon OpenSearch Service provided full-text search capabilities for quick and relevant search results across user-generated content.
For audio content transcription and translation, Amazon Bedrock was used to convert speech into text, supporting multiple Indian languages crucial to Arré’s diverse user base. When building its recommendation engine, Arré Voice evaluated many models but ultimately chose Amazon Titan for its cost-effectiveness and better integration with AWS infrastructure, accelerating development timelines. Machine learning models trained using AWS Batch enabled continuous optimization of the recommendation engine’s algorithms. By combining Astra DB’s real-time data handling with the scalability of AWS services, Arré Voice created a recommendation engine capable of delivering personalized content to millions of users in real time across multiple languages
Outcome | Generating 30 Million Recommendations a Month
Arré Voice successfully migrated to its microservices architecture in just 45 days, incorporating powerful AWS services such as Amazon ECS, Amazon DynamoDB, Amazon Neptune, and Amazon OpenSearch Service. If the company had used its second-best database option, it would have required an additional 20–25 days of development time due to a lack of integrations—and an increased spend of $8,000–$10,000. As such, the company saved about 45 percent on development costs. The rapid migration set the foundation for a robust, high-performing recommendation engine. As a result of these improvements, the new infrastructure achieved remarkable gains in performance. Latency for critical services dropped by 90 percent, from over 1000 milliseconds to consistently around 100 milliseconds, vastly improving response times and enhancing the user experience. The move to DataStax Astra DB played a pivotal role in executing real-time queries execution and handling of vast data volumes. The platform's cloud-native design simplified deployment and minimized operational complexities, while its developer tools allowed the engineering team to continuously refine and optimize the recommendation algorithms.
Within 30 days of implementation, the platform was delivering over 30 million personalized suggestions, significantly enhancing user engagement with more relevant content. According to Shabeer Muhammed, senior software engineer at Arré Voice, “Our recommendation engine surpassed our expectations, delivering 30 million personalized suggestions each month. DataStax Astra DB and AWS were pivotal in our success, helping us achieve our goal within a remarkably short timeframe.” The improved infrastructure not only enhanced performance but also achieved cost-effectiveness compared to previous database solutions. With the platform now operating at 99.9 percent availability, Arré Voice exceeded its SLA and SLO requirements to gain the reliability and scalability necessary for future growth.
