Saymico & Parpple: dating apps, shipped end to end
Two consumer dating apps taken from design to the App Store and Google Play: matching, AI photo verification, real-time chat and a moderation console.
- Photo trust score
- 99.4%
- Store approval
- 1st Submission
- Stack
- React Native / Node.js / WebSockets / AI verification
- IP
- Transferred to the client
Saymico and Parpple are two consumer dating apps targeting the Indian market. The central challenge in dating apps is trust. Users will not engage if they suspect fake profiles, and a single bad experience (harassment, catfishing) can permanently damage an app's reputation. Both apps needed to launch on iOS and Android from a single codebase and be approved by app stores on first submission.
Constraints
- iOS App Store and Google Play have strict guidelines for dating apps, and moderation tools are required
- Photo verification had to be accurate enough to prevent fake profiles without adding significant friction
- Real-time chat at consumer scale requires a robust WebSocket architecture
- A moderation console for the client's team was a hard requirement before launch
AGA built both apps using React Native (Expo) sharing a single codebase with platform-specific adjustments for iOS and Android. An AI photo verification system compares profile photos against a liveness selfie to compute a trust score. Real-time chat is delivered over WebSockets with end-to-end message delivery guarantees. A web-based moderation console allows the client team to review reports, suspend accounts, and manage the content policy.
How the system is structured.
Simplified. Implementation details are withheld where they are commercially sensitive.
- 01
React Native app
Single codebase (Expo) compiling to iOS and Android
- 02
User onboarding
Phone OTP verification, profile creation, photo upload
- 03
AI liveness check
Selfie compared to the profile photo with computer vision to compute a trust score
- 04
Matching engine
Location-based candidate generation, preference filtering, match scoring
- 05
WebSocket chat
Node.js WebSocket server with persistent connections, delivery receipts and typing indicators
- 06
Media storage
Profile photos and chat media stored in S3 with CDN distribution
- 07
Report pipeline
User report → review queue → moderation console → action
- 08
Moderation console
Web dashboard for client team: view reports, review profiles, take action
What makes it hold up.
- Trust scores computed once at upload and cached, not re-run on every view
- WebSocket connections stateless-compatible: server state in Redis so any backend node handles any connection
- Message delivery guaranteed by client-side retry with server-side deduplication on message ID
- Photo moderation runs asynchronously: the upload completes immediately and approval doesn't block
- All profile photos pass through automated NSFW detection before becoming visible
- Match candidates pre-computed in background jobs, so swipes read from cache instead of live queries
Key decisions, and why we made them.
- 01
Expo (React Native) instead of separate native codebases
Two separate native codebases doubles engineering effort and creates divergence risk. Expo provides native performance for the features required (camera, push notifications, location) while allowing a single development team to own both apps.
- 02
Custom AI trust scoring instead of third-party verification service
Third-party face verification services charge per verification and introduce a dependency on an external API's uptime and pricing. AGA built the trust scoring pipeline using open-source computer vision models, which lowered the per-check cost and kept full control over the algorithm.
- 03
WebSocket chat instead of push notifications for real-time delivery
Push notifications are delayed, unreliable on some Android OEMs, and cannot carry full message content for privacy reasons. WebSockets provide sub-second delivery for active users and fall back to push notifications for background delivery.
More shipped work and related services.
- 01CityXmart: a 41-second WhatsApp checkoutCustomers browse the catalogue, fill a cart and pay by UPI without leaving WhatsApp. The store owner gets an alert the moment an order is paid.Conversational Commerce
- 02Jaipur garment ERP: six logbooks, one dashboardA garment manufacturer in Jaipur replaced six paper logbooks with one system for job cards, inventory and GST-ready invoicing, used on 10-inch tablets on the shop floor.Shop-Floor & Management
- 03Build servicesWeb apps, mobile apps, SaaS products and online stores, engineered to production standard.Services
- 04Ready-made solutionsLaunch-ready app platforms you can brand as your own.Solutions
- 05Technology StackThe tools we use and why each one earns its place.Engineering
Need something similar?
Describe your project and we'll scope it, design the architecture and send a fixed-price quote within 24 hours.
- Written scope
- Within 24 hours
- Pricing
- Fixed per milestone
- Ownership
- 100% of code and IP