← Back to Project

AI & Knowledge Platform

AI-power video knowledge platform — HackAIthon 2025

Kato (Quan Ngo)
Kato (Quan Ngo)
3 min read
--

SuperVault — HackAIthon 2025SuperVault — HackAIthon 2025

“AI-First isn’t just a theme — it’s a mindset for how we build, learn, and share knowledge.” — FPT HackAIthon 2025 Committee

🏆 Overview

Project: SuperVault — AI-Powered Video Knowledge Assistant Achievement: 🥈 Second Prize, FPT HackAIthon 2025 Event Theme: AI in Learning & Knowledge Management Date: August 2025 Team: GHC.AICodeCure

🎯 Problem Statement

Every year, FPT Software produces thousands of hours of internal training videos. While this content is rich with knowledge, much of it remains unstructured and underutilized — difficult to search, summarize, or reuse effectively.

The challenge for HackAIthon 2025:

Build an AI assistant that transforms video content into structured, searchable knowledge — enabling smarter learning and faster insight discovery.

🚀 Our Solution — SuperVault

SuperVault is an AI-powered video knowledge vault that automatically processes, analyzes, and indexes video content into a semantic search platform.

Using Azure AI Video Indexer and our custom-built Onyx AI engine, we transformed raw videos into actionable insights — including transcripts, sentiment analysis, visual recognition, and keyword-based semantic search.

Core Capabilities

  • Audio & Visual Intelligence: Speech-to-text, language detection, speaker identification, OCR, and object recognition.
  • AI-Powered Insights: Sentiment analysis, summarization, topic extraction, and timeline-based transcript generation.
  • Semantic Search & Indexing: Smart chunking, BM25 ranking, and hybrid vector-text search via Vespa.
  • Integration Layer: RESTful APIs connecting internal learning systems (Confluence, LMS, S3, etc.).

🧠 System Architecture

  • SuperVault Core: Upload management, authentication, progress tracking, and workflow orchestration.
  • Azure AI Video Indexer: Audio transcription, visual analysis, and structured insights extraction.
  • Onyx AI Platform: AI-driven semantic indexing, hybrid vector search, and custom agent integration.
  • Persistent Layer: Multi-cloud storage using AWS S3, Azure Blob, Vespa (Vector DB), PostgreSQL, and Redis.
  • Data Flow Pipeline:
    1. Upload → Video to Blob Storage
    2. Analyze → Extract audio/video insights
    3. Process → Summarize, visualize, and transform
    4. Index → Ingest content into Onyx Search
    5. Search → Query semantically via API or UI

SuperVault Architecture DiagramSuperVault Architecture Diagram

💡 Key Highlights

  • End-to-End AI Integration — fully automated video analysis pipeline.
  • 15x Faster Knowledge Discovery — semantic search replaces manual browsing.
  • Cross-Platform Ready — unified knowledge hub for multiple data sources.
  • Enterprise Scalability — designed for internal deployment across FSOFT divisions.

SuperVault DemoSuperVault Demo

🧭 Reflection

SuperVault wasn’t just a Hackathon project — it was a vision for the future of knowledge management at FPT Software. It showcased how AI can transform passive content into active intelligence, bridging the gap between data, peopl

Comments