Graph Connect
Six Degrees
A social platform that thinks in relationships. It keeps 50 thousand connections in a graph database and answers who knows who in two hops, with profiles and content living in their own store alongside.
- Python
- FastAPI
- TypeScript
- Neo4j
- MongoDB
- Docker
- AWS
Problem
Social features — friend recommendations, mutual connections, follower graphs — are awkward to express in a relational or document store. I wanted to build a platform that uses the right datastore for each job and still behaves like one coherent backend.
Approach
- Polyglot persistence: Neo4j holds the social graph (50K+ relationships); MongoDB holds profiles and content documents.
- FastAPI microservices front both stores behind one API surface, with JWT-based auth.
- Cypher 2-hop traversals power “people you may know” recommendations; full-text search runs across profiles and posts.
- Three-layer test suite: pytest for units, Locust for load, Selenium for end-to-end flows — all containerized with Docker and deployed on AWS.
Result
A working graph-backed social platform demonstrating multi-datastore architecture, graph query design, and a test pyramid that actually exercises the system under load.