To meet these needs, a comprehensive backend solution was developed using AWS services, with a focus on modularity, automation, and AI integration. At the heart of the solution is AWS Bedrock, which provides access to powerful foundation models such as Amazon Titan and AI21 Labs. These models enable the chatbot to process natural language queries and generate personalized, context-aware responses.

On the Bedrock side, specialized tooling is used to handle data fetching and context retrieval. Before invoking the foundation model, user queries are embedded using a vector embedding model (also available through Bedrock or other ML services). These embeddings are compared against a vector database—Amazon OpenSearch—which contains pre-indexed representations of the astrology knowledge base. This architecture allows the system to dynamically fetch relevant context from the content corpus and inject it into the prompt sent to the LLM. While model fine-tuning is not required, Bedrock supports orchestration techniques such as retrieval-augmented generation (RAG),where contextual snippets are fetched and used to guide the model’s response generation.

The application backend itself is deployed using AWS Lambda for scalable and event-driven compute. Amazon API Gateway serves as the interface layer for client requests, while Amazon Cognito handles user identity and access management. The backend logic is distributed across private subnets within a Virtual Private Cloud (VPC) that spans two availability zones, ensuring high availability and network isolation.

For data storage, the solution uses Amazon RDS (PostgreSQL) for structured records and Amazon S3 for storing datasets used in embeddings and knowledge base updates. Redis (Amazon ElastiCache) is used to cache frequent queries and reduce response latency.

Deployment and monitoring are fully automated using AWS CodePipeline, CloudFormation, and CloudWatch, giving the team the ability to quickly release changes, monitor performance, and respond to any issues in production.