Hybrid machine translation architecture combining rule-based Prolog logic with neural NLLB-200 models on AWS, deployed across public presentation, private application, and data layers within a VPC. Users access a Django-powered Bootstrap UI through an ALB and WAF, routing requests to an Orchestrator Service that coordinates Prolog rule evaluation and AI-driven translation. MongoDB stores lexical, morphological, and grammatical knowledge; SQLite manages app metadata; S3 backs up models and datasets; CloudWatch and CloudTrail provide observability. This architecture demonstrates how to balance symbolic reasoning (Prolog) with deep learning (Transformers) for linguistically-aware translation, solving the cold-start and interpretability challenges of pure neural approaches. Fork this diagram on Diagrams.so to customize subnets, add multi-region failover, or integrate additional translation engines. The setup includes automated model download, dataset import via Pandas, and credential management through Secrets Manager and KMS.