Placement Interview Preparation Assistant

Overview

This project is a Retrieval-Augmented Generation (RAG) based system that generates company and role specific technical interview questions. The system retrieves context from a local knowledge base and uses a language model to generate structured interview question sets with short answers.

Features

• Generates interview questions tailored to a specific company and role.
• Retrieval-Augmented Generation pipeline to ensure responses stay grounded in stored context.
• Local inference (no external API dependency).
• Simple Gradio-based user interface.

System Flow

1. User enters a query containing the company and/or role.
2. System performs semantic search in a vector database (ChromaDB).
3. Top matching context chunks are retrieved.
4. Language model generates interview questions based on retrieved context.
5. Output is displayed in the UI.

Tech Stack

Component	Tool
Embeddings	SentenceTransformer
Vector Store	ChromaDB
Retrieval Framework	LangChain
Language Model	Phi Mini (local)
UI	Gradio
Environment	Python, GPU backend (Vast.ai or local CUDA)

Setup Instructions

1. Install Dependencies

pip install -r requirements.txt

2. Ensure Models Are Available Locally

models/
 ├── phi_mini/                  # LLM directory
 └── embeddings/                # SentenceTransformer embedding model

3. Start the Application

python gradio_ui.py

4. Access the UI

If running locally:

http://localhost:7860

If running on cloud (e.g., Vast.ai), use the mapped public port:

http://<instance-ip>:<public-port>

Directory Structure

project/
 ├── models/
 │   ├── phi_mini/
 │   └── embeddings/
 ├── chroma/                       # Vector DB storage
 ├── gradio_ui.py                  # Gradio interface
 ├── rag_pipeline.py               # Retrieval and generation logic
 ├── README.md
 └── requirements.txt

Future Enhancements

• Add aptitude and HR question modules.
• Add role-wise dataset expansion.
• Improve UI layout and presentation styling.
• Optional web deployment and user authentication.