UyooNotes

Overview

A note service in which AI reads notes as context rather than as files. UyooLabs’ own product, designed and built end to end — from the knowledge model through the agent pipeline to the web application.

Core Problem: Search Is Not the Same as Remembering

Notes accumulate faster than anyone can organize them, and keyword search returns only what the user already knew to look for. The challenge is not storage but surfacing the right note at the relevant moment, in a form the user recognizes as their own thinking rather than a generic answer.

Design approach: keep the source note as the unit of truth, retrieve by meaning rather than string match, and always accompany answers with pointers back to what the user actually wrote — so a wrong answer is visibly wrong rather than quietly plausible.

Key Contributions

• Designed the full personal knowledge system — note model, context construction, retrieval, and recommendation unified as one pipeline rather than bolted-on search
• Built the generative AI agent pipeline — multi-model routing that assigns volume to lightweight models and judgment to stronger ones
• Implemented meaning-based retrieval and recommendation — RAG over a vector store, tuned for personal-scale corpora where precise recall matters more than breadth
• Architected the web application end to end — data model, authentication, storage, and client

Tech Stack

• AI: OpenAI GPT, Gemini, Claude, RAG, Vector Database
• Frontend: Next.js, TypeScript, Tailwind CSS
• Backend: Supabase, PostgreSQL