SwiftSwipe
Tap–search–decide: instantly pick the optimal credit card & grow rewards intelligently.
Dec 2024 - Apr 2025Flutter, NestJS, PostgreSQL, RAG
Retrieval Augmented Generation (RAG)FinTechMobile
Case Study
Card choice in 3 taps
Architect a system that recommends the best card at the moment of payment.
Problem
People juggle multiple cards but leave rewards unused because rules are opaque and decisions happen under time pressure.
Objectives
What success looked like
- 3 taps from intent to choice
Protocol
Recommendation flow
- 1Choose merchant (search/pick) and enter amount
- 2Resolve merchant → normalize name, map to MCC (RAG fallback)
- 3Fetch benefits/constraints for user's saved cards
- 4Score each card (category match, caps, exclusions, promo windows)
- 5Return a single best recommendation with a short reason
- 6Log decision + anonymized outcome for analytics
Outcomes
What shipped
- +18–24% average reward capture in test cohort
- Single, confident recommendation reduces decision time to seconds