We build software.We make it scale.

WorkStack is a multi-user Scrum tool — workspaces, projects, sprints and a drag-and-drop Kanban board — built as a functioning product and then taken through the same path a real small production service follows: containerized, deployed to Kubernetes, provisioned on GCP with Terraform, delivered through CI/CD, and watched with real observability.

What was built

A REST API under /api/v1 backs workspaces, projects, sprints and tasks, with cookie-based JWT sessions and tenant isolation enforced server-side — every workspace-scoped resource checks membership before returning anything, and a sprint's tenant boundary is derived transitively through its project rather than stored directly. The board itself is drag-and-drop, with optimistic moves reconciled against the API response.

Why it was built

Most infrastructure-focused projects either wrap a toy app in heavy tooling, or build real infrastructure with nothing meaningful running on it. WorkStack does both properly: a genuinely usable product, deployed and operated the way a small engineering team would actually run it — with the application and the platform underneath it treated as two separate, deliberately built layers.

Architecture

React and TypeScript on the frontend, FastAPI and PostgreSQL on the backend, with SQLAlchemy for the ORM and Alembic for migrations. Application concerns and infrastructure concerns are kept strictly separate — the product layer is frozen once it reaches MVP, while the platform underneath is built up one deliberate stage at a time: Docker, then Kubernetes locally, then Terraform, then GKE, then Helm, CI, GitOps and observability.

Infrastructure & deployment

Terraform provisions the GCP foundation — VPC, GKE (Standard, zonal, Spot-node autoscaling), Cloud SQL on a private IP, Artifact Registry and Secret Manager. The application ships as a single Helm chart. GitHub Actions runs backend and frontend tests, a Trivy scan and a Docker build, then pushes immutable git-SHA-tagged images to Artifact Registry — CI never touches the cluster directly. Argo CD watches a separate GitOps repository and syncs the new image into GKE.

Observability

Prometheus and Grafana are provisioned as infrastructure via Terraform, kept outside the application's own Helm chart, and track the four golden signals — traffic, latency, errors and saturation — across both the backend and the cluster. The Grafana dashboard is defined and shipped as code, not clicked together by hand.