Abstract

The project is a personal portfolio chatbot built with Streamlit, LangChain, and Cerebras GPT-OSS-120B. It acts as a conversational agent that represents Ramachandra Udupa, answering questions about his work, projects, services, and experience. The chatbot uses Retrieval-Augmented Generation (RAG) to pull accurate information from curated sources such as a PDF resume and personal websites, ensuring contextually correct responses while maintaining the persona of Ramachandra.

Introduction

Personal branding is critical in technology careers. Static resumes and portfolios limit interaction and personalization. This project introduces an interactive AI-powered chatbot that serves as a living portfolio, providing a conversational interface for recruiters, clients, or collaborators to explore professional details dynamically. The chatbot is designed to emulate the voice and style of Ramachandra, making the interaction more engaging and authentic.

Objectives

1. To build an AI chatbot that embodies Ramachandra’s persona.
2. To enable recruiters and clients to interactively explore portfolio information.
3. To ensure responses remain truthful, concise, and context-specific.
4. To integrate resume data and portfolio websites into a retrieval pipeline.
5. To deliver a smooth user experience through a custom Streamlit interface.

System Overview

PortfolioBot is a Retrieval-Augmented Generation (RAG) bot designed to provide intelligent answers about a user’s professional background, projects, and experience by leveraging data from publicly available websites and personal documents such as a resume. The system combines retrieval-based search and generative language modeling to deliver accurate and context-aware responses.

How It Works – Step by Step

1. Data Collection

   • PortfolioBot first fetches information from multiple sources, including:

     • Personal websites (LinkedIn, portfolio sites, GitHub, etc.)
     • Resume documents (PDF, DOCX, or TXT formats)

   • This ensures that the system has a comprehensive view of the user’s professional profile.

2. Data Preprocessing and Chunking

   • The raw content is split into smaller text chunks for efficient retrieval.
   • Chunking ensures that the RAG model can retrieve precise, context-specific answers instead of scanning large documents inefficiently.

3. Vectorization

   • Each text chunk is converted into a vector embedding using the All-MiniLM embedding model.
   • Embeddings encode semantic meaning, allowing similar content to be matched even if the wording is different.

4. Storing in Chroma DB

   • The vector embeddings are stored in Chroma DB, a high-performance vector database.
   • Chroma enables fast similarity searches, allowing the system to retrieve the most relevant chunks in real time.

5. Retrieval-Augmented Generation (RAG)

   • When a user asks a question, PortfolioBot performs the following:

     1. Converts the query into an embedding.
     2. Searches the Chroma DB for the most relevant chunks using similarity search.
     3. Passes the retrieved chunks along with the original query to a language model for answer generation.

   • This ensures that the answers are grounded in the user’s real data rather than being generated from generic knowledge.

6. Response Delivery

   • The generated response is returned to the user via a Streamlit web interface, providing an interactive and user-friendly experience.
     

Installation

Follow these steps to install and run PortfolioBot:

1. Clone the repository

git clone https://github.com/Ramachandra-2k96/RAG_Project
cd RAG_Project

2. Set up Python environment
   Make sure you have Python 3.12 installed. You can check with:

python3 --version

3. Create a virtual environment (recommended)

python3 -m venv venv

Activate the virtual environment:

• Windows:

venv\Scripts\activate

• macOS / Linux:

source venv/bin/activate

4. Install dependencies

pip install -r requirements.txt

5. Run the bot

streamlit run main.py

Results

The system provides consistent, natural, and concise answers while remaining grounded in verified data. Tool usage ensures retrieval is only triggered when necessary. The bot never fabricates unrelated information, aligning strictly with professional portfolio details.

here is the preview link : https://portfolio-rag.streamlit.app/

Conclusion

The project demonstrates how AI can transform a portfolio into an interactive, conversational agent. By merging LLMs with RAG, Ramachandra’s professional identity is represented authentically and dynamically. This approach improves accessibility, engagement, and personalization in professional networking, offering a forward-looking alternative to resumes and static portfolio websites.