Case Study 1 of 7

Productivity / AI•2025•Live in Production

Memrizor

Intelligent Knowledge Assistant & Model Context Protocol (MCP) Vector Retrieval

My Role

Full-Stack Engineer & Product Architect

Timeline

4 Months

Team

Solo Project

Primary Stack

React, Express, Node.js

Visit Live App View Repository

Key Verified Outcomes & Metrics

300+

Resources Indexed

Active bookmarks & articles processed

<120ms

Search Latency

Semantic vector retrieval speed

45%

Search Time Saved

Reduction in time to find saved references

99.9%

Uptime

Reliable cloud availability

Project Overview & The Core Problem

Eliminate the digital chaos of hoarding bookmarks and tabs across multiple devices. Memrizor acts as an intelligent knowledge assistant that automatically summarizes, categorizes, and semantically links technical knowledge using Model Context Protocol (MCP) and LLMs.

My Responsibilities & Contributions

Architected the end-to-end full-stack application using React, Node.js, and MongoDB Atlas.
Integrated OpenAI API and Model Context Protocol (MCP) servers for autonomous knowledge tagging and semantic queries.
Engineered an asynchronous job queue to scrape metadata and generate embeddings without blocking the client thread.
Designed a clean, keyboard-friendly UI with responsive search and instant collections.

Technical Architecture

System Design & Component Flow

1. Frontend Client

React SPA with Tailwind CSS & optimistic UI state updates

2. Backend & APIs

Node.js & Express REST Gateway with rate limiting & JWT auth

3. Database / AI Tooling

MongoDB Atlas with compound indices and collection trees

Data Flow Pipeline

1. Capture

Browser extension / UI captures URL & user annotations

2. Ingest

Express worker retrieves article DOM and extracts key entities

3. Vectorize

OpenAI generates semantic embedding stored in MongoDB

4. Query

MCP server enables natural language queries across the second brain

Engineering Challenges & Implemented Solutions

Building in the real world means encountering unexpected edge cases, latency bottlenecks, and technical trade-offs.

Challenge 01

Asynchronous Web Scraping at Scale

Third-party websites often block or delay headless scraping, causing high latency on single-threaded Express endpoints.

Solution Implemented

Background Worker Queue

Decoupled scraping from user interactions using an asynchronous worker queue that parses OpenGraph tags and creates vector embeddings in the background.

Challenge 02

Semantic Vector Search vs Exact Matching

Balancing precise keyword matching with fuzzy semantic understanding for bookmarks without inflating database costs.

Solution Implemented

Hybrid Search Pipeline

Combined MongoDB text index filters with lightweight OpenAI embedding vectors for two-stage candidate retrieval.

Technologies Utilized in Memrizor

ReactExpressNode.jsTailwind CSSMongoDBMCPOpenAI

Core Capabilities Shipped

Instant 1-click web bookmarking with auto-extracted title, favicon, and preview card
AI-driven automated tagging and thematic collection clustering
Natural language semantic search powered by MCP and OpenAI
Clean markdown note editor with syntax highlighting for code snippets

Final Outcome

Memrizor achieved over 300+ actively indexed resources with sub-120ms retrieval times, dramatically streamlining technical knowledge management for early adopters.