To address these challenges, a serverless and highly scalable architecture was implemented using AWS-managed services. Real-time communication was established through Amazon API Gateway using WebSocket protocols, allowing persistent connections with mobile and web clients. A custom Lambda authorizer validated user authentication tokens issued by Amazon Cognito and enforced single-session-per-user logic. Sessions were cached in Redis for quick access and replicated in DynamoDB as a fallback store. This enabled both performance and consistency, even during cache failures or concurrent logins.

User location updates were streamed every 15 seconds through WebSocket and processed by a dedicated Lambda function. These locations were indexed in Redis using geospatial data structures, enabling real-time lookups for nearby users and alerts. Alert data was fetched from multiple public RSS sources and third-party providers using Step Functions that executed every minute. These workflows triggered Lambda functions to ingest, filter, deduplicate, and store alerts in Redis with TTLs to ensure timely cleanup. A secondary Lambda regularly cleaned expired keys from Redis, and another archival Lambda converted near-expiry alerts into Parquet format for storage in Amazon S3, preserving historical records for compliance and analytics.

For scalability, the architecture used AWS CDK (in TypeScript) to define and deploy infrastructure-as-code. CI/CD pipelines were built using GitHub Actions, enabling fast and reliable deployments across development and production environments. Regions were isolated to avoid data collisions and maintain fault tolerance. Monitoring and logging were handled through Amazon CloudWatch, and critical system failures were sent to internal teams via Slack with deduplication to avoid noise. This gave the team full visibility into the health of the platform.