From GCP to AWS: Unlocking a Broader AI Model Ecosystem for Global Scale
Protagona partnered with an AI-driven SaaS company to re-platform its iOS application backend from Google Cloud to AWS, integrating Amazon Bedrock and establishing a production-grade hybrid cloud architecture built for scale.
Industry
Startups & Software
Teams & Services
Cloud Architecture, Cloud Engineering, Engagement Management
Tech & Tools
Amazon Bedrock, AWS Fargate, Amazon ECR, Amazon RDS, Amazon Neptune, Amazon VPC, AWS WAF, Elastic Load Balancing, Amazon Route 53, AWS IAM, Amazon CloudFormation
Key Data Points
The Vision
Built to deliver intelligent narrative experiences through an iOS application, this AI-driven platform had outgrown its Google Cloud backend. Running on a single foundation model provider constrained the product roadmap. The decision to shift the backend to AWS was deliberate — designed to leverage Amazon Bedrock's expanding model catalog and AWS's global regional footprint, not simply swap hosting providers.
The Goal
Protagona was engaged to achieve three concrete objectives: migrate the containerized backend and relational database from GCP to AWS, integrate Amazon Bedrock with appropriate model access and guardrails, and deliver a production-grade AWS environment the client team could operate independently. The GCP-resident frontend would remain in place, requiring seamless cross-cloud communication from day one.
The Challenge
The core complexity was not rehosting — it was re-platforming an active AI backend across two cloud providers while keeping a GCP-resident frontend in continuous operation. The architecture had to bridge GCP and AWS networking without disrupting iOS end users, while replacing GCP-native AI workflows with Amazon Bedrock integrations that matched prior functionality. The database migration required a snapshot-and-restore approach to move a live relational database to Amazon RDS with data integrity intact, alongside deploying a net-new Amazon Neptune graph database. Every component — networking, compute, database, and AI integration — had to be production-ready from day one.
The Solution
Protagona began with a structured discovery phase, mapping the existing GCP environment across networking, container registry, compute, and database layers. That assessment drove a purpose-built AWS architecture grounded in managed services — chosen to minimize operational burden without requiring a dedicated infrastructure team. Compute was built on AWS Fargate and Amazon ECR with service and task configurations mirroring the GCP containers. Networking was constructed from scratch: a custom VPC with defined routes, endpoints, Route 53 DNS, load balancers, and AWS WAF for perimeter security — designed to accept traffic from the GCP frontend while enforcing consistent security policies across the hybrid boundary.
The relational database migrated to Amazon RDS via snapshot-restore, and Amazon Neptune was deployed for graph-based AI data structures. Bedrock integration included full model access configuration and guardrails appropriate for a customer-facing product. The engagement closed with complete knowledge transfer — diagrams, implementation guides, and all infrastructure-as-code artifacts.
.png)