Cloud Architecture
Design and operation of scalable cloud-native infrastructure on AWS, Google Cloud and bare metal Kubernetes.
What's included
- Amazon Web Services (EKS, RDS, S3, CloudFront)
- Google Cloud Platform (GCP)
- Managed and bare metal Kubernetes
- Infrastructure as Code (Terraform)
- High availability and disaster recovery
- Cloud cost optimization
I design and operate resilient, cost-effective cloud infrastructure. From microservices architecture to the daily running of production Kubernetes clusters.
The scale I work at
I run Kubernetes clusters on EKS with more than 60 nodes and around 500 workloads in production, integrated with the AWS services around them: RDS for relational databases, ElastiCache for caching, CloudFront for delivery, S3 for storage, SES for email, KMS for key management, and DocumentDB.
In parallel I run bare metal Kubernetes clusters on Hetzner, with dozens of nodes and hundreds of workloads. These are two different sets of problems, and having both in daily operation is what makes it possible to recommend one over the other with actual grounds.
Managed or bare metal
On a managed service such as EKS the control plane is the provider's responsibility: upgrades, availability and integration with the rest of the cloud services come included, and you pay for that convenience both in fees and in constraints.
On bare metal the per-node cost drops substantially and control is total, but the control plane, networking, persistent storage and upgrade plan become yours. It is a sustainable choice when someone is genuinely looking after that cluster.
There is no universally correct answer. The deciding factor is usually the ratio between infrastructure spend and the cost of the skills needed to run it in-house.
Certification and method
I am Google Cloud certified as an Associate Cloud Engineer, certified in 2024.
The infrastructure I deliver is described as code with Terraform rather than clicked together in a console: it is reproducible, changes go through review, and disaster recovery is a documented procedure instead of a memory exercise.
Frequently asked questions
Do you only work on Google Cloud?
No. The certification is Google, but day-to-day operations cover AWS, Google Cloud and bare metal Kubernetes on Hetzner. Platform choice follows load and budget, not preference.
Can you reduce our current cloud spend?
It is one of the most requested activities. It starts with an analysis of what is actually running and with which allocated resources: the savings are usually in oversized workloads, forgotten storage and traffic that should be served by a CDN.
Do you also handle operations, or only the build?
Both. Several clusters I designed are still under my operational management, which is why architecture decisions account for whoever has to maintain them.