FinTech · Engineering, no AI
Payment routing for MENA, LATAM and Southeast Asia
A single payment provider stopped working once the client sold in new regions. An orchestration layer now routes each transaction by rule, and reroutes it when a provider fails.
- Client
- Recurring billing and subscription management for media and entertainment brands
- Segment
- FinTech, payments
- Regions
- MENA, LATAM, Southeast Asia
- Built on
- Primer orchestration, with Adyen, Yuno and regional providers
- Primer
- Adyen
- Yuno
- Node.js
- AWS Lambda
- Amazon SQS
- Amazon SNS
- Amazon S3
- AWS CDK
- SST
- AWS Vault
Before
What was true before
Expansion into MENA, LATAM and Southeast Asia exposed the limits of a setup with one payment provider. Approval rates and revenue fell, partly from data formatting mismatches and partly from missing local coverage. Each new provider took months of development and QA, routing was static, and an outage at the provider could stop every transaction.
- Static routing, with no way to send a transaction to the provider most likely to approve it
- Months of work to add each new provider
- One provider outage could halt all payments
- Problems reported by users before monitoring caught them
After
What we did
An orchestration layer on Primer connects several providers behind one API. A rule-based engine routes and retries each transaction from its metadata, a second processor takes over when one fails, and dashboards show approval rates, decline reasons and latency per provider.
Build log
How it was built
- 01
Built an orchestration layer on Primer connecting Adyen, Yuno and regional providers
- 02
Designed an API-first, event-driven microservices architecture integrated with Primer and the client SRM system
- 03
Built a rule-based engine for routing and retries driven by transaction metadata
- 04
Added multi-processor redundancy that reroutes transactions automatically during a provider outage
- 05
Built dashboards for approval rates, decline reasons and latency per provider
- 06
Set up infrastructure as code with SST, credentials in AWS Vault, and CI/CD with automated tests
- 07
Added real-time monitoring and alerting in place of user-reported issues
Measured
Results
As reported in the published case.
- faster processor integration, from months to weeks
- 80%
- requests per second handled under load
- 10K+
- control over payment workflows through orchestration
- 100%
- A 20 to 30% boost in approval rates in key regions is projected on the published case, not yet measured
- The architecture targets 99.99% uptime
This engagement did not include AI. It is here as evidence of the engineering that AI work sits on.
Engagement
Where this work sits
Next step
Yours would be the next one.
Tell us where it breaks, and you get a written answer on what breaks first and what it costs.