Powering the Future of Learning: Building a Scalable AWS Home for Cursuri
The Client
Cursuri-AI is Romania’s first AI-native online learning platform, operated by GALIAN Software Development SRL (Cluj-Napoca, Romania). It delivers premium Romanian-language courses across two tracks: IT & engineering and AI for business – one platform where both developers and business professionals can turn AI into a real advantage at work.
What sets Cursuri-AI apart is that AI isn’t just the subject – it’s built into the learning experience itself: a 24/7 AI virtual teacher, auto-generated quizzes and summaries, and hands-on practical exercises, all in native Romanian and backed by human-authored editorial quality.
- Industry:Embedded Software Products
- Company Size:2-10
- Country:Romania
Challenges
- Production-grade infrastructure with a lean team – The team needed a reliable, auto-scaling cloud platform for a content-heavy web application and an AI-assisted content pipeline, without the overhead of building out a large in-house cloud operations team.
- Managing AI workload elasticity – With the unpredictable nature of AI-assisted processing, the infrastructure required a flexible architecture that could handle sudden spikes in demand without compromising performance.
- EU data residency and compliance – Serving Romanian and EU learners meant navigating strict GDPR alignment and EU AI Act readiness, requiring all hosting and data processing to remain strictly within the EU control 1 region.
- Disciplined early-stage cloud spend – To validate the product without overspending, it was critical to make sound architectural decisions from day one and maximize on AWS credit runway.
- Balancing infrastructure demands with rapid validation – As a founder led company building a modern AI learning platform, the goal was to establish a high-performing architecture that supported compliance and scaling without burning through resources.
Building a modern AI learning platform as a lean, founder-led company came with specific AWS and infrastructure challenges, requiring external expertise to confidently scale and secure their environment without limiting future growth.
Solutions
Cloudvisor started with a cost review and then worked closely with the team to define practical ways to apply Savings Plans effectively based on their real usage.
Architectural guidance for an AWS-native platform
A modern, containerised infrastructure was designed on AWS Fargate, ensuring a clean separation between the public learner-facing experience and background AI content workloads.
Maximising early-stage cloud funding
We at Cloudvisor were able to provide support to secure AWS Activate credits, giving the team meaningful budget headroom during their critical MVP and early growth phases.
Implementation of security best practices
An extensive review was conducted on identity and secrets management, establishing secure CI/CD pipelines via AWS-native federation to completely eliminate long-lived cloud credentials.
Optimising infrastructure for AI workload scaling
EKS node AMIs and container configurations were tailored to efficiently handle the heavy processing demands of the background AI pipelines without delaying user-facing services.
Ongoing partner-level account management
After the initial deployment, Cloudvisor continues to work closely with the team to provide billing oversight, support tier alignment, and right-sizing guidance as the platform continues to grow.
AWS Services Used
- Amazon RDS
- Amazon ElastiCache for Redis
- Amazon SQS
- AWS Secrets Manager
- GitHub Actions integrated with AWS IAM OIDC
- Amazon CloudWatch
- Amazon SageMaker
- Amazon VPC
- AWS IAM
Results
Talkpal came to Cloudvisor with a fast-growing platform and uncertainty around how to optimise AWS spending without limiting future growth. What followed was a structured cost overhaul — and the start of a continuous FinOps partnership that keeps spending efficient as Talkpal scales to serve learners worldwide.
High-Availability DeploymentZero-downtime deployments and automated rollbacks were implemented to ensure continuous platform availability.
- 1.5 Years of Infrastructure CoveredInfrastructure costs covered for an estimated 1 to 1.5 years based on current growth pace.
-
18% Cost ReductionA shift to AWS SQS for message handling delivered an 18% reduction in cloud costs.
- $10K in AWS Credits SecuredCloudvisor helped secure $10K in AWS Activate credits, with an additional $15K in progress.
- Elevated Security PostureHardened security by centralizing secrets, enforcing least-privilege access, and eliminating long-lived CI/CD keys.
- Streamlined Compliance FrameworkFull EU data residency was established from day one to support strict GDPR and EU AI Act compliance.
