Interlinked

Gemini said Interlinked: Privacy-First Social Discovery The Vision: Moving social media from "Engagement" to "Encounter" by bridging the loneliness gap with predictive matchmaking.

Comment

Inspiration

The "Why" Behind the Stack Our goal was to build a social ecosystem that delivers deep human resonance without the ethical pitfalls of centralized data harvesting. We chose a Hybrid-Edge AI architecture to decouple "Personal Signal" from "Cloud Storage."

  1. Systems Architecture Frontend: Built with React Native (Expo) to ensure a cross-platform, high-performance UI capable of handling real-time video tracking and native haptics.

Database & Auth: Supabase (PostgreSQL) serves as our relational backbone. We leveraged the pgvector extension to handle high-dimensional similarity searches, enabling us to match users based on mathematical "trajectories" rather than static strings.

Edge AI Engine: Instead of heavy cloud-side processing, we implemented a custom LocalInferenceService. Using MurmurHash3 feature hashing, we project 40+ behavioral signals into a normalized 384-dimension vector directly on the device.

  1. Security & Data Privacy (The DevOps Moat) Our DevOps pipeline is designed with Zero-Knowledge principles:

On-Device Processing: Raw behavioral DNA (dwell time, re-watch counts, caption sentiment) is processed in a transient local state and stored in encrypted AsyncStorage. It never leaves the device.

Anonymous Vector Syncing: Our sync service only pushes the final float array (the Vibe Vector) to the server. Even if our database were breached, the data is mathematically anonymous and cannot be reverse-engineered into the user's private scrolling history.

Auditability: We utilized PatriotAI during the development phase to audit our algorithm's governance, ensuring our "Black Box" logic remains ethical and free of reinforcement bias.

  1. Scalability & Resilience Real-time Communication: Using Supabase Realtime, we implemented a low-latency messaging and notification system that scales horizontally.

CI/CD Pipeline: We utilized a modular service-oriented architecture, allowing us to swap the local hashing logic for full ONNX native models in the future without disrupting the core Social Graph.

  1. Tech Stack Summary Languages: TypeScript, SQL

Frameworks: React Native, Expo, NativeWind

Infrastructure: Supabase (PostgreSQL + pgvector), AsyncStorage

AI/ML: Google Gemini API (Semantic Analysis), MurmurHash3 (Local Hashing)

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Interlinked

Built With

  • expo-haptics
  • languages:-typescript
  • nativewind-(tailwind-css)
  • realtime)-databases:-postgresql-with-pgvector-extension-apis:-google-gemini-api-(semantic-trajectory-analysis)-libraries:-expo-video
  • sql-(pl/pgsql)-frameworks:-react-native-(expo-sdk-50+)-cloud-services:-supabase-(auth
  • storage
Share this project:

Updates