AI knowledge base · Web + Android

Social Brain.

A place for everything you save. A cross-platform knowledge base with an AI assistant grounded in your own sources.

Visit live product
ARTICLENOTEBOOKMARKKNOWLEDGE INDEXpgvector / hybrid searchGROUNDED ANSWER[1] [2] [3] sourcesEXTRACT → ANALYZE → EMBEDARCHITECTURE ILLUSTRATION
DOCUMENTED OUTCOMEWeb + AndroidOne shared codebase
TECHNOLOGIES

TypeScript · Next.js 15 · React Native · PostgreSQL / pgvector · Claude API · BullMQ · Redis · Turborepo

The problem

Saved content needs to become useful knowledge. That means extracting, organizing, and retrieving it without making people wait for each AI operation to finish.

The engineering

01

Built and deployed a Next.js web app and an Expo React Native client from a single Turborepo, sharing TypeScript types, Zod validation schemas, and AI logic.

02

Designed an asynchronous pipeline on BullMQ and Redis: extract, analyze, embed, complete. Workers summarize and tag content through the Claude API, with exponential-backoff retries.

03

Combined HNSW-indexed cosine similarity with PostgreSQL full-text search and reranking. The RAG assistant grounds responses in indexed items and includes inline source citations.

The outcome

Delivered a knowledge base across web and Android, with shared business logic and background processing that keeps user requests from blocking on LLM inference.

PROJECT SUMMARY BASED ON MY RÉSUMÉ. DIAGRAMS ILLUSTRATE THE ARCHITECTURE; THEY ARE NOT PRODUCT SCREENSHOTS.
NEXT PROJECT / 02Enterprise RAG