AI-Powered Excel Mock Interviewer

Design Document & Strategy

1. Solution Architecture

Core Approach: Conversational AI agent with structured interview flow and intelligent evaluation system.

Technology Stack:

• Backend: Python + FastAPI (lightweight, async, easy deployment)
• LLM: OpenAI GPT-4 (best reasoning for complex Excel evaluation)
• Frontend: Streamlit (rapid prototyping, built-in chat interface)
• State Management: Pydantic models + session storage
• Deployment: Docker + cloud platform (AWS/Render)

2. System Components

A. Interview Flow Manager

• Structured conversation states (intro → questions → evaluation → summary)
• Dynamic question selection based on difficulty progression
• Context-aware follow-up questions

B. Excel Knowledge Evaluator

• Multi-dimensional scoring: Technical accuracy, approach quality, efficiency
• Question bank covering: Formulas, Data Analysis, Pivot Tables, VBA, Best Practices
• Adaptive difficulty based on performance

C. State Management System

• Session persistence across conversation turns
• Performance tracking and scoring accumulation
• Interview transcript logging

Excel Mock Interviewer

Automated system to assess a candidate's Excel skills via a multi-turn interview simulation.

Features

• Structured interview flow (intro, questions, summary)
• Dynamic, LLM-generated questions
• Intelligent answer evaluation
• Agentic interviewer behavior
• Constructive feedback and report card
• Interactive audio interface:
  • Text-to-speech for questions and feedback
  • Voice input for answers (Web Speech API)
  • Both text and voice input available simultaneously
• Simple React frontend for Q&A

Tech Stack

• Backend: FastAPI (Python)
• LLM: OpenAI GPT-4o-mini (or Mistral 7B)
• Frontend: React
• Storage: In-memory (no DB)

How to Run

Backend (API)

1. Install Python dependencies:

   pip install -r requirements.txt

2. Start FastAPI server:

   uvicorn app:app --reload

Frontend (React)

1. Go to frontend folder:

   cd frontend

2. Install dependencies:

   npm install

3. Start React app:

   npm start

API Endpoints

• POST /start — Start interview, get intro and first question
• POST /answer — Submit answer, get feedback and next question
• GET /summary — Get interview summary

Stretch Goals

• Speech-to-text (Whisper API)
• Audio output (TTS)
• Performance score summary

---

Built by GitHub Copilot