Next Project

Topic Modelling

Esplanade Project Preview
D E B B Y

Follow

Virtual Veteran: AI-Powered Organisation Memory

I built Virtual Veteran, an AI chatbot that acts as a virtual laojiao ("old hand"), keeping the wisdom of past colleagues alive by tapping on their emails. Selected for presentation at PST Ideathon 2024, this innovative solution addresses the challenge of preserving organizational knowledge when experienced professionals move on.

Task

Built an AI-powered chatbot that preserves and makes accessible organizational knowledge from email archives.

  • Role

    Full Stack Developer

  • Organization

    Self-initiated

  • Date

    2024

  • Tools

    Python, Flask, OpenAI GPT-4, Gradio, TF-IDF

Status: Selected for PST Ideathon 2024 🚀

Virtual Veteran Project

THE CHALLENGE


Problem Statement

The idea struck when I realized how much wisdom and know-how we lose whenever seasoned professionals move on to other roles. Important institutional knowledge often disappears, leaving new team members struggling to find answers to questions like "Who handled last year's big event?" or "What was the outcome of that important meeting that happened before I joined?"

SOLUTION


Technical Approach

Virtual Veteran leverages cutting-edge technologies to preserve and make accessible organizational knowledge:

  • OpenAI GPT-4 for natural language processing and response generation
  • Gradio for creating an intuitive user interface
  • TF-IDF for finding the most relevant email responses
  • Flask for robust backend server operations
  • Python for seamless data handling and processing

Key Features:

  • Users can simply drag and drop their email archive, with automatic processing of PDF attachments
  • Questions are processed against the archive to fetch relevant information
  • Responses include email subject titles and dates for context
  • The system provides concise, accurate answers powered by historical communications

DEMO


Product Demo

REFLECTION


Learning Journey

Though the product remains a prototype, developing it from scratch was an invaluable learning experience. It challenged me to apply my Flask development skills in new ways and marked my first implementation of Retrieval-Augmented Generation (RAG). The project's selection for presentation at PST Ideathon 2024 validated its potential impact on organizational knowledge management.

Share this post