The problem
Helpicon matches people with vetted local helpers. The platform had to take payments in two systems, split commissions automatically, keep chat private between parties and deploy safely — with a small team.
What I built
- 01Integrated Stripe and Swish (Nordic mobile payments) with automatic commission splits, subscription tiers and webhook handlers for payments, subscriptions and refunds.
- 02Cut Twilio messaging costs by 90% with a shared Conversations service, multi-tenant isolation and S3 attachments.
- 03Made the API 2–3x faster — under 200 ms — with field projection, caching and composite indexes; database queries under 50 ms.
- 04Ran ECS with auto-scaling (2 to 10 tasks at 70% CPU) and Multi-AZ RDS PostgreSQL with automated backups.
- 05Deployed the payment microservice on EKS with Jenkins and Kaniko across three AWS accounts (DevOps, staging, production), all in Terraform, with Prisma migrations on every deploy.
Screens


