OmniMind AI — Multimodal Knowledge Engine

Inspiration

In today's fast-paced tech landscape, developers, researchers, and knowledge workers are flooded with information across fragmented formats: 50-page PDF research papers, 2-hour recorded team calls, slide decks, and developer notes. Traditional search engines fail because they cannot link insights across different media types.

We built OmniMind AI to solve information overload by creating a unified cognitive engine that maps text, vision, and audio into an interactive visual knowledge graph.

What it does

OmniMind AI is an Enterprise Multimodal Knowledge & Insights Engine. Key features include:

• 📂 Multimodal Knowledge Vault: Drag & drop PDFs, DOCX, MP3/WAV Audio, and Images for instant vector embedding and executive summaries.
• 💬 Grounded Neural RAG Chat: Interactive conversational assistant with an active document context selector and inline source citations.
• 🕸️ Interactive SVG Knowledge Graph: A visual 2D node map showing cross-modal concept connections with a live Concept Inspector panel.
• 🎙️ Audio & Voice Studio: Synchronized audio waveform player with line-by-line speaker transcript highlighting and synthetic neural voice generation.
• 📊 Enterprise SaaS Workspaces: Toggle between R&D Team, Product Strategy, and Personal Vaults with SOC2 compliance status indicators.

How we built it

• Frontend & Core: React 18 bundled with Vite for ultra-fast response times.
• Styling & Theme: Custom Cyber Glassmorphism system built with Vanilla CSS (Plus Jakarta Sans & JetBrains Mono fonts).
• Icons & Visuals: lucide-react icons and custom SVG canvas graph rendering.
• Interactive FX: canvas-confetti for celebration micro-interactions.

Challenges we faced

• Context Grounding: Ensuring the AI assistant strictly distinguishes between querying a single specific PDF vs. synthesizing answers across the entire multi-file repository.
• Visual Graph Layout: Calculating dynamic coordinate placement for SVG nodes and edges without overlap.
• Audio Transcript Sync: Aligning audio timestamp cues with interactive UI highlight states.

Accomplishments that we're proud of

• Building a polished, high-performance web app that feels like a real-world enterprise product rather than a simple hackathon prototype.
• Seamlessly integrating document vaulting, visual graph exploration, and speech synthesis under one cohesive dark-mode UI.

What we learned

• How to structure multimodal vector embeddings for fast cross-document Q&A.
• Designing intuitive glassmorphic UIs that present complex AI metrics simply.

What's next for OmniMind AI

• Real-time live microphone stream transcription and instant node graph generation during meetings.
• Browser extensions for 1-click web page indexing into OmniMind Vault.

Built With

• canvas-confetti
• css3
• html5
• javascript
• lucide-icons
• rag
• react
• svg
• vector-embeddings
• vite