About This Architecture
Real-time voice AI platform on AWS ap-south-1 (Mumbai) combining Route 53, CloudFront CDN, AWS WAF, and Shield for global edge delivery and DDoS protection. Traffic flows through dual ALBs and NLBs across two availability zones, routing to ECS microservices (API Gateway, Conversation Manager, LLM Orchestrator, Voice Agent Worker, LiveKit WebRTC server) with CPU/memory auto-scaling. Asynchronous workloads fan out via SQS queues (stt-processing, llm-processing, background-task, retry) with queue-depth scaling, while ElastiCache Redis and RDS PostgreSQL (primary/standby) provide session state and persistent storage across private subnets. CloudWatch, X-Ray, CodePipeline, and IAM complete the observability and CI/CD backbone. Fork this diagram to customize VPC CIDR ranges, add additional regions, or adjust SQS queue configurations for your voice AI workload. This architecture demonstrates multi-AZ resilience, serverless messaging patterns, and managed database failover essential for production voice applications.