Cloud Engineering Intern
I spent four weeks on Ejada's Cloud Build track. It started with Oracle Cloud Infrastructure and Terraform, moved through secure networking and Oracle Kubernetes Engine (OKE), and ended with a team-built e-commerce delivery pipeline using Helm, GitHub Actions and ArgoCD. The first three weeks were individual labs in my own OCI compartment, and week four was a group capstone in a shared compartment and GitHub organisation.
Cloud Build cohort and mentors in Cairo
What I worked on
Individual cloud engineering labs
- Built OCI networking, compute, block storage, and File Storage manually, then reproduced and extended the environment with Terraform
- Designed a two-tier OCI network with a public Application Load Balancer, private compute, NAT, Bastion access, and internal NFS storage
- Created reusable Terraform modules for an OKE environment and deployed a stateful Nginx workload with OCI block-volume storage
- Verified each environment through runtime evidence and removed cloud resources in dependency order to control shared programme costs
Group A capstone
- Contributed to an e-commerce system spanning four Node.js microservices, a Next.js frontend, MongoDB, and OCI-hosted Kubernetes infrastructure
- Worked with Terraform, Helm, OCIR, GitHub Actions, NGINX Ingress, cert-manager, and ArgoCD across the delivery path
- Documented integration fixes and production-readiness gaps, including authorisation, payment confirmation, secret management, database resilience, and the absence of automated tests
Completed Work
OCI infrastructure and secure network design
Progressed from a basic public VCN and compute deployment to a two-tier architecture where the load balancer was the only public application entry point and the instance remained private.
- Built with
- OCI, Terraform, Linux, Bastion, File Storage
- Contribution
- Built the first three weeks' environments individually, first exploring the resources in the OCI Console and then expressing them as repeatable Terraform configurations.
- Outcome
- Verified compute, storage, load-balancer routing, private access, and controlled teardown in the assigned OCI compartment.
Reusable OKE deployment
An Oracle Kubernetes Engine environment built from reusable subnet and cluster modules, with VCN-native pod networking, a managed worker pool, public load-balancer exposure, and persistent OCI block storage.
- Built with
- OKE, Kubernetes, Terraform modules, Nginx, OCI Block Volumes
- Contribution
- Provisioned the environment, deployed and debugged the workload, destroyed it to control cost, and recreated it from the same code the following day.
- Outcome
- A browser-reachable stateful application reproduced in two independent infrastructure cycles.
E-commerce modernisation capstone
I was part of Group A. We took an e-commerce project that one of our teammates had already built and modernised it as a team: four Node.js services, a Next.js frontend, MongoDB, Terraform-provisioned OCI networking and OKE, Helm packaging, and GitOps delivery through ArgoCD.
- Built with
- Node.js, Next.js, MongoDB, Helm, GitHub Actions, ArgoCD
- Contribution
- Wrote the project documentation for the capstone: the existing and target architecture, application and infrastructure components, testing, CI/CD and GitOps flows, integration troubleshooting, security review, known limitations and the handover notes.
- Outcome
- A source-to-cluster pipeline where an application push builds multi-architecture images, updates the Helm state and triggers ArgoCD to reconcile, with the architecture, test evidence, problems and remaining risks written up in one document.
Technical Stack
Cloud and infrastructure
Oracle Cloud Infrastructure, Terraform, Linux administration, Cloud networking, Cost governance
Containers and delivery
Docker, Kubernetes, OKE, Helm, GitHub Actions, ArgoCD, GitOps
OCI services
VCN, Compute, Load Balancer, Bastion, File Storage, Block Volumes, Object Storage, OCIR
Week by week
Week 1: OCI and Terraform foundations
Built compute, networking, and storage manually, then automated the same dependency chain and diagnosed a compartment-scope authorisation failure.
Week 2: Secure two-tier networking
Moved the application instance into a private subnet behind a public load balancer, with NAT, Bastion, and File Storage supporting controlled access.
Week 3: Kubernetes on OKE
Introduced reusable Terraform modules, deployed a persistent workload, and resolved worker-image, container-image, and storage-aware rollout failures.
Week 4: Team capstone and GitOps
Integrated application repositories, OCI infrastructure, Helm, CI workflows, ingress, TLS automation, and ArgoCD into one collaborative delivery system.
Problems along the way
The first Terraform apply returned an OCI authorisation-style error.
Authentication was valid, but the configuration targeted the shared parent compartment rather than the assigned leaf compartment where resource management was permitted.
Verified the authentication chain and requested scope separately, then passed the correct assigned compartment OCID to the resources.
The subsequent Terraform apply completed successfully.
A Kubernetes replacement pod could not attach the existing ReadWriteOnce block volume while the previous pod still held it.
The default rolling update strategy conflicted with the storage access mode of the single-replica stateful workload.
Changed the Deployment strategy to Recreate so the old pod released the volume before its replacement started.
The rollout completed and the storage-backed application returned to a healthy state.
The capstone crossed several architecture and dependency boundaries during integration.
ARM64 builds, registry authentication, probe timing, OCI load-balancer annotations, remote Terraform state, TLS, and workflow order each failed at a different layer.
Used runtime evidence to isolate each layer, then applied targeted fixes including multi-architecture images, permanent OCI auth tokens, adjusted probes, injected subnet metadata, remote Object Storage state, cert-manager, and corrected workflow ordering.
The team established an automated application and infrastructure delivery path while recording its remaining limitations.
Results
- Programme length
- 4 weeks (Aug – Sep 2026)
- Individual lab phases
- 3
- Capstone application repositories
- 5
- Primary cloud region
- Saudi Arabia West (Jeddah)
- The documentation keeps my individual work separate from the group capstone and lists the known security, resilience and completeness gaps, so the demo isn't presented as production-ready.