AI-Staff Augmentation: AI-Powered Care Assistant Robot
TinyBots needed a stable, scalable engineering team to power Tessa across 150+ European care providers, InApps's IT Staff Augmentation delivered the cross-device platform behind the robot.
Digital Banking Transformation
NovaPay Financial Group is a mid-market neobank operating across Southeast Asia with 2M+ registered users. Founded in 2017, they grew rapidly but their legacy monolithic platform became a bottleneck: slow releases, high downtime, and a poor mobile experience were causing customer churn.
The competitive neobank landscape demanded a modern mobile-first experience. NovaPay's engineering team lacked the bandwidth and expertise to execute a full platform overhaul while keeping production stable. They needed a trusted technology partner to lead the rebuild.
Project goals
- Rebuild core banking platform on cloud-native microservices architecture
- Deliver iOS and Android apps with feature parity within 36 weeks
- Integrate AI-powered fraud detection with <50ms transaction scoring
- Achieve 99.9% uptime SLA post-launch
- Reduce operational costs by at least 40%
Client requirements
Across each key area, these requirements shaped every architectural and product decision we made.
Mobile Application
NovaPay needed a native mobile presence on both iOS and Android, with a seamless user experience that could compete with top-tier fintech apps in the region.
- Native iOS and Android apps with shared business logic
- Biometric authentication (Face ID, fingerprint)
- Real-time transaction feeds and push notifications
- Offline mode for balance viewing
Core Banking
The legacy monolith had to be replaced without downtime. NovaPay required a modern, event-driven backend that could scale independently per service.
- Microservices architecture with clear service boundaries
- Event-driven transaction processing
- Multi-currency support (USD, VND, SGD, THB)
- Automated reconciliation engine
AI & Security
Regulators and customers demanded real-time fraud protection and full data security compliance across all transaction flows.
- Real-time fraud scoring under 50ms per transaction
- Anomaly detection with weekly model retraining
- PCI-DSS Level 1 compliance
- End-to-end encryption for all data in transit and at rest
Infrastructure
The platform had to be resilient enough to handle rapid user growth and regional expansion without engineering intervention.
- 99.9% uptime SLA with zero-downtime deployments
- Auto-scaling to handle 10x peak traffic
- Multi-region failover (Singapore + Ho Chi Minh City)
- Full observability: logs, metrics, distributed tracing
Business challenges
The core obstacles that shaped every decision we made in rebuilding this platform.
Legacy Monolith
A 7-year-old Java monolith with no clear service boundaries. Any change risked breaking unrelated features, making iterative modernisation nearly impossible.
Zero Mobile Experience
No native mobile app and a mobile web score of 28/100 on Google Lighthouse, resulting in a 64% bounce rate on mobile devices.
Fraud & Security Gaps
Manual fraud review caused 4-6 hour delays in flagging suspicious transactions. Regulators demanded a real-time automated detection system.
Scalability Ceiling
Peak-hour traffic caused database connection exhaustion. The system could not scale horizontally without creating race conditions in the transaction ledger.
Our solutions
Each solution addresses a specific pain point while contributing to a unified, future-proof platform, from infrastructure to mobile to AI, every layer rebuilt with scale, security, and speed in mind.
Strangler Fig Migration
Incrementally replaced the monolith using the strangler fig pattern, routing traffic to new microservices behind a facade API gateway while keeping legacy services alive during transition.
React Native Cross-Platform App
Built a single React Native codebase sharing ~80% of business logic across iOS and Android, with native modules for biometric auth and performance-critical screens.
ML Fraud Detection Pipeline
Deployed a dual-model system on AWS SageMaker: a gradient-boosted tree for known patterns and an autoencoder for anomaly detection, scoring every transaction in under 50ms.
Event-Driven Architecture
Replaced synchronous API chains with Apache Kafka event streams, enabling horizontal scaling and eliminating the transaction race conditions that caused data inconsistencies.
Technology stack
Leveraging high-expertise engineering and battle-tested technologies, the InApps team delivered a scalable, production-ready solution built for long-term growth.
React Native, TypeScript
Node.js, NestJS, Apache Kafka, GraphQL
AWS ECS
PostgreSQL, Redis, DynamoDB
The experts behind it
Every successful delivery starts with the right people. InApps assembled a dedicated, cross-functional team structured for speed, accountability, and long-term partnership.
Project Manager
Owns delivery timeline, stakeholder communication, and risk management across all 6 phases.
Solution Architect
Designed the microservices architecture, API gateway, and event-driven Kafka topology.
Backend Engineers
Built core banking services: auth, accounts, transactions, notifications — each independently deployable.
Mobile Engineers
Delivered the React Native app for iOS & Android with shared business logic and native biometric modules.
DevOps / Cloud
Provisioned AWS infrastructure with Terraform, set up CI/CD, observability, and zero-downtime deployments.
QA Engineers
End-to-end testing, penetration testing, and PCI-DSS compliance validation across all services.
What we delivered
Key outcomes
- Mobile app reached 4.8 rating on both App Store and Google Play within 60 days
- Transaction processing latency reduced from 2.3s to 180ms average
- Zero critical incidents in first 6 months post-launch
- Engineering team onboarded to new platform in 3 weeks with full documentation
- NovaPay expanded to 2 new markets within 4 months, using same platform
Within 90 days of launch, NovaPay recorded their highest-ever monthly active user count. The new fraud detection system caught 94% of fraudulent transactions automatically, reducing manual review workload by 85% and saving an estimated $2.3M annually in fraud losses.
Ready to build something that matters?
Tell us what you are building and the constraints you are working under, and we will come back with a team shape that fits.