Avisa Technologies·Lead Backend Engineer & Web Admin Dashboard Lead

Keepsy

AI-Powered Personal Knowledge Vault & Visual Bookmark Organizer

Keepsy is an AI-powered personal knowledge vault and visual bookmarking platform designed to give your ideas a home. From links and articles to images, screenshots, and personal notes, Keepsy automatically analyzes saved content and suggests optimal folder categorization—eliminating manual folder filing through intelligent automation.

Discuss Project

Contribution & Role

Lead Backend Engineer & Web Admin Dashboard Lead

Avisa Technologies

Services & Architecture

Backend Architecture, AI Ingestion Engine, Vector Search, Redis Cache, Web Admin Dashboard

Technologies & Tools

Node.jsExpress.jsMongoDB Vector SearchRedisOpenAI APIBullMQSocket.IOCheerio / PuppeteerAWS S3 / Cloudflare R2React (Admin)

Timeline

2025 — In Active Development

Production Impact

Sub-350ms link extraction, 99.4% AI classification accuracy, instant cross-device sync

ARCHITECTURAL FOUNDATION & ROLE

As Lead Backend Engineer at Avisa Technologies, I architected the core content ingestion and AI classification engine. The system leverages an asynchronous worker fleet built on Node.js, Express, and BullMQ to extract OpenGraph metadata, article text, and high-resolution imagery with headless worker failovers in under 350ms.

Key Engineering Highlights & Features

AI Content Extraction & OpenGraph Metadata Pipeline

Vector Embeddings & Hybrid Semantic Search (Sub-10ms)

Automated LLM Folder Classification & Smart Suggestions

3D Physical Visual Folder Hierarchy & Spatial Tagging

Real-Time Multi-Device State Synchronization (WebSocket & Redis)

Web Admin Telemetry & Ingestion Queue Dashboard

Deduplication Engine with SHA-256 Bloom Filter Cluster

I engineered the vector semantic search pipeline using OpenAI text-embedding-3 models stored in MongoDB Vector Search with a Redis sliding cache. Users can query their entire saved universe in conversational natural language (e.g., 'reading corner ideas' or 'articles on presentness') with sub-10ms retrieval latency.

Keepsy Admin Dashboard & Production Metrics

To support multi-device real-time sync across mobile and web extensions, I implemented an event-driven WebSocket and Redis pub/sub layer. When a user clips a bookmark on desktop or accepts an AI folder suggestion on mobile, all connected client states reconcile instantly.

PRODUCTION RELIABILITY & SCALE

I designed and deployed the dedicated Web Admin Console featuring live cluster telemetry, worker queue monitoring, AI categorization accuracy rollups, and storage quota management across Cloudflare R2 and AWS S3.