Production defects per release fell 80%; regression went from 3 days to 40 minutes.
Quality and Delivery: QA and Delivery Automation
My role: Engineering Manager
Engineering
How I build.
I started in engineering, moved into product, and now lead both. The mobile apps I’m responsible for move real money, so I care as much about how they’re built and shipped as about what they do.
How hands-on I am today: actively involved in the details, without losing sight of the bigger picture.
Mobile architecture
Feature modules with one source of truth.
Each feature (transfers, sub-accounts and approvals, the partner plug-in) is a self-contained module with its own state machine, so a screen can only show a state the money is actually in. Permissions are loaded on the device for display, but every money-moving request is re-checked by the server.
Reliability lives in the networking layer: idempotency keys on every send, retry with back-off, and safe resumption when the app closes mid-transfer.
Presentation
Flutter UI. Screens render from the user’s role and the feature’s current state; nothing is shown that the user can’t do.
Feature modules
- Transfers
- Sub-accounts and approvals
- Bills
- Cards
- QR Pay
- Partner plug-in
Self-contained, each with its own state machine and tests.
State and domain
Explicit states (draft → confirmed → processing → successful, failed, reversed). State management: Riverpod, BLoC / Cubit, Provider, GetX.
Networking
Idempotency keys on every money request, retry with back-off, certificate pinning.
Offline
Safe resumption if the app closes mid-transfer; cached reads.
Secure storage
Tokens and credentials in the platform keystore; PIN or biometric to confirm.
Server re-checks role, account and limit on every money-moving request. The client is never trusted.
Stack
- Flutter
- Kotlin/Swift
- React Native
QA automation
Quality built in, not checked at the end.
Before, each release took 3 days of manual regression. Now most coverage runs on every pull request, and the slow end-to-end suite guards the journeys where money moves: login, send, receive, bills and cards. Each product’s engineers own their tests; there is no separate QA handoff.
Production defects per release, indexed
6 months before
6 months after
Delivery automation
From commit to store, with gates that stop themselves.
-
Pull request
Under 20 min
Lint, type checks, unit tests, smoke end-to-end suite.
All green to merge
-
Merge and build
Every merge
Built for staging or development and shipped internally for testing.
Internal build passes
-
Release candidate
Weekly
Automated store builds, release notes, release checklist.
Critical-path tests pass
-
Staged rollout
About 5 days
1% → 10% → 50% → 100%, monitored in Crashlytics and Sentry.
Crash-free ≥ 99.5%
- 1%
- 10%
- 50%
- 100%
Gate between every step: crash-free sessions ≥ 99.5% in Crashlytics and Sentry, or automatic halt.
- Release cadence
- Bi-weekly → weekly, by month 3. My team’s work.
- Rollback
- A crash or error spike during the staged rollout halts it automatically, before most users see the release. The POS terminal app rolls out by partner cohort with automatic rollback.
- Feature flags
- Unfinished work ships behind flags, so we release weekly regardless of feature size. The cost: flag clean-up is ongoing work.
Tools
- GitHub Actions
- Codemagic
- Crashlytics
- Sentry
- Bitrise
Engineering practices
The habits that keep it working.
Definition of done
Tests are part of done. A feature isn’t finished until its money paths are covered.
Observability
Crash-free sessions are tracked in Crashlytics and Sentry; they rose from 98.7% to 99.7% over 6 months, and they gate every rollout at 99.5%.
Code review
Improves code quality and consistency while fostering knowledge sharing and collective ownership. Creates opportunities for engineers to learn from one another, catch defects earlier, and reduce avoidable rework.
Incident handling
Strengthened the team’s ability to respond to and recover from production issues. Established clearer ownership, encouraged root-cause analysis, and shifted the culture from simply fixing bugs to preventing recurring incidents and improving system reliability.
The full story
Contact
Have a product that needs to ship well?
I’m open to global roles leading product, mobile engineering, or both. Email is the fastest way to reach me.