From Azure to AWS: Sort A Brick's Migration Story

Back

The Client

Sort A Brick (operated by BR1CK, UAB) is a Lithuania-based startup leveraging AI-powered computer vision and automated sorting technology to analyze, clean, and restore mixed LEGO brick collections into complete, buildable sets. Customers ship their loose bricks to the company, which identifies restorable sets, sources missing pieces, and returns them fully organized.

By automating the time-consuming process of brick identification, Sort A Brick enables the reuse of existing toys and promotes sustainability by actively reducing plastic waste for eco-conscious consumers across Western Europe.

Industry:Consumer Services / Circular Economy
Company Size:~13 employees
Country:Lithuania

Overall, great communication, always very clear, and technical expertise are top notch.

Aurimas Slapšys

Co-Founder

Challenges

Sort A Brick was previously running a highly fragmented, data-intensive architecture distributed across Microsoft Azure (virtual machines and blob storage), DigitalOcean, local on-premise machines, and external GPU providers like RunPod and VastAI for machine learning. This presented several critical business limitations:

High operational costs – Cloud storage was a massive financial drain, with Azure Blob Storage alone accounting for ~$7,000–$9,000 of their total $12,000–$16,000 monthly infrastructure spend.
Cross-cloud performance bottlenecks – Processing approximately 80 TB of outbound data weekly across disparate cloud and GPU providers created severe inefficiencies and risked the strict sub-0.5 second latency requirements needed for real-time ML image inference.
Security & reliability gaps – Infrastructure lacked enterprise-grade controls, relying on publicly exposed VMs and manual credential management in .env files.
Costly downtime risks – With only minimal, manual monitoring (file-based logs and Slack alerts), platform instability threatened their strict uptime requirements, carrying an estimated €300/hour revenue impact during peak operating hours.

Solutions

Cloudvisor executed a lift-and-shift migration and infrastructure consolidation from Azure and third-party providers to AWS, adapting the data-transfer approach mid-project after initial cost modeling underestimated Azure egress fees.

Phase 1: AWS Foundation & Security Baseline

Provisioned a secure, 3-tier multi-AZ VPC in the Frankfurt region using Terraform infrastructure-as-code.
Deployed a comprehensive security and observability baseline—including AWS WAF, Amazon GuardDuty, AWS CloudTrail, and AWS Secrets Manager—eliminating public VM exposure by routing all access through private subnets and AWS Systems Manager.

Phase 2: Scalable Compute & Machine Learning Consolidation

Rehosted application workloads to Amazon EC2 Auto Scaling Groups utilizing Graviton-based processors (ARM) to optimize price-performance without altering the customer’s existing deployment workflows.
Validated a proof-of-concept on Amazon SageMaker with an initial 6 TB dataset, enabling on-demand, pay-per-use ML model training ahead of full production rollout.

Phase 3: High-Volume Data Migration & Database Modernization

Migrated the operational PostgreSQL database to Amazon RDS, providing automated backups, encryption at rest, and simplified maintenance.
Executed an AWS DataSync transfer to Amazon S3, aggressively cleaning and reducing raw data volume with the customer after an Azure-egress cost estimate came in higher than planned.

Phase 4: Traffic Cutover & Knowledge Transfer

Safely transitioned live production traffic to AWS behind an Application Load Balancer using automated AWS Certificate Manager (ACM) SSL certificates.
Delivered complete Terraform code repositories and operational runbooks, ensuring the engineering team maintained full self-service control.

AWS Services Used

Amazon EC2 Auto Scaling Groups
Amazon S3
Amazon RDS for PostgreSQL
Amazon SageMaker

Application Load Balancer (ALB) & AWS WAF
AWS DataSync
AWS Secrets Manager
Amazon GuardDuty

Results

The migration empowered Sort A Brick with a unified, secure AWS setup that dramatically cut cloud spend, consolidated machine learning workflows, and secured operations against costly downtime.