ASend with Zelleto a contact, with a memopendingsentsent
Deposit
Zelle P2Pmobile check depositzero PCI data on device
Overview
Overview
First Tech Federal Credit Union serves members who are primarily tech workers at Google, Microsoft, HP, and Intel.
I delivered the React Native mobile banking app across the full money-movement surface area: mobile check deposit, Zelle P2P, bill pay, internal and external transfers, and card controls — with banking-grade security and a test suite to match.
The challenge
Members compared the app to Venmo, not to other credit unions.
The legacy app was slow, couldn't do P2P, and had a deposit flow that failed too often. Every money-movement bug is a support call, and every security gap is a regulatory incident — the bar was 'zero incidents' and 'feels like a fintech app' at the same time.
Fintech barcompared against Venmo
Money movementevery bug is a support call
Securityzero incidents was the bar
The solution
Native bank SDKs behind one React Native app.
Rebuilt the hot flows on React Native with native bridges wrapping the bank SDKs (MI-SNAP, Zelle, Plaid, Card.io) so members got native-quality experiences from a single codebase. Layered optimistic UI over money movement with explicit rollback on failure, and gated every release behind a 95% test coverage threshold. Session security stayed out of the way with biometric unlock and device binding.
Native SDK bridgesOptimistic UIReal rollback
Architecture
Native SDKs, hardened auth.
A React Native app on top of a hardened native auth stack, connected to bank rails via tokenized third-party SDKs. Every sensitive flow bridges to native code, and no card data ever touches the JavaScript thread.
Client
React Native
Redux
Styled Components
Optimistic UI with rollback
Auth
Biometric unlock
Keychain / Keystore
Session tokens
Device binding
Native Bridges
MI-SNAP (check deposit)
Zelle SDK (P2P)
Plaid (linking)
Card.io (scan + tokenize)
Backend
mTLS to core banking
Tokenized transactions
Zero PCI data on device
Audit-grade logging
One request, start to finish
Member opens app and biometric unlock restores session
Balances and transactions stream with optimistic UI
Check deposit auto-captures via MI-SNAP and tokenizes server-side
Zelle P2P sends with contact lookup and memo
Every action logs to audit with device + session binding
Decisions
Five calls and why I made them.
Call 01
React Native with native bridges for bank SDKs
MI-SNAP, Zelle, Plaid, and Card.io all ship native. Wrapping them as native modules kept the app cross-platform without losing vendor support — the alternative was either two fully native apps or pinning to an old SDK version forever.
Call 02
Optimistic UI for money movement, with explicit rollback
Banking apps feel slow because every action waits for a server round-trip. Optimistic updates with a rollback state machine cut perceived latency in half without corrupting the balance — which is non-negotiable in money software.
Call 03
95% test coverage as a release gate
Money bugs don't get a soft launch. Jest + React Native Testing Library + Detox + Appium on a CI gate meant no regression could reach production, and made the app safe to iterate on fast.
Call 04
Biometric unlock + device binding, not passwords
Passwords are phishable and friction-heavy. Tying the session to a device, requiring biometric unlock, and revoking on device change raised the attack bar and dropped login time from seconds to one tap.
Call 05
Zero PCI data on the device, ever
Card numbers go scanner → tokenization SDK → server without ever sitting on the JS thread or in state. Staying out of PCI scope on the device eliminated an entire compliance surface and made the app safer to iterate on.
Reflections
What I took away.
The benchmark isn't other banks — it's Venmo
Our members were tech workers at Google, Microsoft, Intel. They didn't compare the app to other credit unions; they compared it to Cash App. That reframed every decision: P2P had to feel instant, deposits had to work first try, and a 2-second tap-to-balance was unacceptable. Knowing who you're actually competing with — not who your industry thinks you're competing with — is the cheapest positioning exercise a product team can do.
Optimistic UI belongs in money software, but only with a real rollback
Banking apps felt slow because every tap waited for a server round-trip. Optimistic updates cut perceived latency in half — but only because we treated rollback as a first-class state, not an edge case. The half-finished version of this pattern is how balances get corrupted. Do it all the way or don't do it.
Keeping PCI off the device is more valuable than any control you can add
Card scans went straight from the scanner into a tokenization SDK and onto the server — card numbers never touched the JS thread or our Redux store. Staying out of PCI scope on the device eliminated an entire compliance surface and made the rest of the app safer to iterate on. The cheapest security control is the one where the data doesn't exist in the first place.
Outcomes
What shipped.
SecurityZeroSecurity incidents since launch
P2P Volume$2M+Monthly Zelle transactions
Test Coverage95%Enforced on every release
Stack
Technology
Frontend
React Native
Redux
React Navigation
Styled Components
Banking SDKs
MI-SNAP
Zelle SDK
Plaid
Card.io
Security
Biometric unlock
Keychain / Keystore
Device binding
mTLS
Testing & Analytics
Jest
React Native Testing Library
Detox
Appium
Firebase
Mixpanel
Crashlytics
Features
What it does
MI-SNAP mobile deposit with auto-capture, edge detection, and retry UX
Zelle P2P with contact lookup, request-to-pay, and memo support — $2M+ monthly volume
Bill pay scheduling with recurring payments and next-payment alerts
Instant card lock/unlock with immediate effect on the auth path
Biometric unlock (Face ID / Touch ID) with device binding and session management
95% test coverage (Jest + Detox + Appium) enforced on every release
Zero security incidents since launch, zero PCI data on device
Analytics and crash monitoring via Firebase, Mixpanel, and Crashlytics