SciVenia & Mubazar Cloud Deployment

Problem

Two TechVenia products — SciVenia (scientific event platform) and Mubazar (musician opportunity marketplace) — needed reliable, secure, multi-region deployments without an internal platform team. Releases were ad-hoc, secrets lived in environment files, and rollback was manual.

Approach

• 01

  Replaced manual SSH deploys with GitLab CI/CD pipelines: lint → test → build → image → deploy.

• 02

  Containerised both apps and ran them on EC2 with an Application Load Balancer and auto-scaling groups across two AWS regions.

• 03

  Moved secrets out of repo into AWS Secrets Manager with per-environment IAM roles for the CI runners.

• 04

  Wired CloudWatch metrics, structured JSON logs to a central aggregator, and added a one-click rollback job that re-deploys the previous image tag.

Outcome

• 01

  Release cycle dropped from ~3 hours of operator time to ~25 minutes of pipeline time — 40% faster end-to-end.

• 02

  Service uptime measured at 99.9% over the first quarter post-migration.

• 03

  Mean time to detect production regressions fell because every release shipped its own dashboards.

• 04

  Onboarding a new engineer to a deploy now takes one short walkthrough instead of a half-day knowledge transfer.

Lessons

• 01

  Treat the pipeline as the product — every deploy is the result of a tested, versioned process, not muscle memory.

• 02

  Push secrets and config into managed services on day one; refactoring them out later costs more than doing it right initially.

• 03

  A rollback button you do not trust is a button you will not press at 3am. Practise the rollback in staging weekly.