Success Story
Clostech migrates its virtual try-on AI from Azure to AWS in 6 weeks, with no service interruption
Clostech is a startup focused on AI applied to e-commerce. Its product lets users try clothes on virtually by uploading a photo and overlaying garments with a proprietary AI model, retrained from open-source models, aiming to reduce returns and increase conversion for digital retailers. Craftech executed its strategic migration from Azure to AWS.
6 weeks
total duration of the AWS migration
Challenge
Clostech ran on Azure with a robust but limited infrastructure: compute-intensive requirements (high-capacity GPU instances to process images in real time), high costs from running 24-core, 98 GB RAM virtual machines, distributed dependencies (frontend and backend on Azure, PostgreSQL databases, Blob Storage), and billing issues that led to temporary account suspensions affecting continuity. The migration had to be completed in just a month and a half, without affecting the AI product's availability during critical sales dates.
Solution
Craftech executed a strategic migration to AWS: Amazon EC2 GPU instances to efficiently process AI workloads, ECS + Docker for backend and AI services in managed containers, Amazon RDS (PostgreSQL) for reliable relational database migration, Amazon S3 + CloudFront for low-latency global image storage and distribution, and Route 53 + ALB for DNS and load balancing with high availability. Access was secured with Pritunl VPN and Identity Center. For observability and automation, CI/CD pipelines were set up with Bitbucket Pipelines, credentials managed with AWS Secrets Manager, and networking optimized (VPC, private subnets, load balancers). A future migration of critical ECS workloads to Amazon EKS was planned.
Results
The migration was completed in six weeks, meeting the deadline ahead of key commercial events. The new infrastructure became more agile and scalable, with dynamic auto scaling for demand spikes and cost optimization from adopting AWS-native services and efficient S3 storage. AI server availability was guaranteed, eliminating billing-related interruption risk, and a solid foundation —with automated pipelines and modern security practices— was left for evolving the AI product and scaling globally.
Stack & technologies
