About the Client

Clostech is a startup focused on applying Artificial Intelligence (AI) to the e-commerce sector.

Its flagship product is an application that allows users to virtually try on clothes. Users upload a personal photo, and the clothing items are overlaid using a proprietary AI model, which was fine-tuned (re-trained) from open-source foundational models.

The company’s goal is to revolutionize the online shopping experience, significantly reducing returns and boosting conversion rates for digital retailers.

🛑 The Challenge

Clostech was operating on Azure but faced critical limitations that compromised efficiency and service continuity:

• Intensive Compute Requirements: The application demanded high-capacity GPU instances to process images in real-time for the virtual try-on feature.
• High Operating Costs: Relying on virtual machines with 24 cores and 98 GB of RAM in Azure proved to be inefficient and difficult to scale cost-effectively.
• Distributed Dependencies: Infrastructure included a frontend and backend on Azure, PostgreSQL databases, and Azure Blob Storage.
• Billing Issues: Temporary account interruptions due to billing problems affected the product’s operational continuity.

The main challenge was to complete a full migration to AWS in just six weeks, ensuring the uninterrupted availability of the core AI service during critical commercial periods.

✅ The Solution

Strategic Migration to AWS

A new infrastructure was designed and implemented on AWS, prioritizing performance, efficiency, and security:

• Amazon EC2 (GPU Instances): Configured specialized instances to efficiently process the demanding AI workloads.
• ECS + Docker: Created Dockerfiles for backend and AI services, deployed in containers managed by Amazon Elastic Container Service (ECS).
• Amazon RDS (PostgreSQL): Performed a reliable and secure migration of the relational databases.
• Amazon S3 + CloudFront: Used for image storage and global, low-latency distribution.
• Route 53 + ALB: Managed DNS and utilized Application Load Balancer (ALB) to ensure high availability.
• Pritunl VPN + AWS Identity Center: Set up secure and centralized access management.

Observability and Automation

• Bitbucket Pipelines: Implemented a Continuous Integration/Continuous Delivery (CI/CD) pipeline for automated deployments to ECS and CloudFront.
• AWS Secrets Manager: Ensured the secure management of credentials and environment parameters.
• Optimized Networking: Configured a secure and efficient architecture using VPC, private subnets, and load balancers.

Future Enhancements and Scalability

• Exposed the AI API using a dedicated Load Balancer.
• Planned the migration of critical workloads from ECS to Amazon EKS (Elastic Kubernetes Service) for greater scalability and flexibility.

🚀 Impact & Outcomes

• Migration completed in six weeks, successfully meeting the deadline before key commercial events.
• Achieved a more agile and scalable infrastructure with dynamic auto-scaling to handle demand peaks.
• Optimized costs by adopting native AWS services and leveraging efficient S3 storage.
• Guaranteed AI server uptime, eliminating the risk of interruptions caused by billing issues.
• Established a solid platform for future AI product evolution, featuring automated pipelines and modern security practices.

Thanks to this migration, Clostech modernized its core infrastructure, significantly improved operational efficiency, and prepared its platform for global scaling, offering its e-commerce clients an innovative, reliable, and high-performance experience.

Key Project Deliverables

• Successful deployment of critical services on Amazon EC2, ECS, and RDS.
• Operational CI/CD pipeline via Bitbucket Pipelines.
• Integration of Amazon CloudFront and S3 for global, low-latency image distribution.
• Reinforced security through AWS Secrets Manager and Identity Center.

Ready to transform your infrastructure and take your business to the next level?

Let’s discuss how Craftech can make it happen for you.