Social Media Backend with ML Matching

Problem

A social product needed a backend that could model relationships and surface relevant people, not just store posts.

Solution

A Node/Express API exposing posts, engagement and social-graph endpoints, with vector embeddings powering interest-based matching between users.

Architecture

Node.js/Express services, PostgreSQL (Neon) for relational + graph data, JWT/RBAC for security, and vector embeddings for similarity scoring.

Highlights

• Posts, engagement and social-graph APIs
• Embeddings-based interest scoring for user matching
• JWT + RBAC access control

Challenges solved

• Modeling a social graph efficiently in Postgres
• Scoring interest similarity with embeddings
• Enforcing role-based access across endpoints

Tech stack