UNL is an AI-powered mock interview coach designed to help students and early-career candidates practice interview responses and receive actionable feedback on both what they say and how they present themselves.

This project combines a React + Vite frontend with a FastAPI backend to deliver:

• Prompt generation from either a built-in prompt bank or pasted job descriptions.
• Real-time posture and eye-contact tracking during interview responses.
• Audio transcription + speaking style analysis.
• LLM-generated interview feedback and scoring.
• A final results dashboard and downloadable report.

Demo video: https://youtu.be/a-7Rl7wk_L0

UNL began as a response to a common frustration in today’s job market: candidates often never hear back or receive meaningful interview feedback. UNL aims to close that gap by simulating interview conditions and returning practical coaching that users can apply immediately.

1. User enters interview preferences (question type + difficulty) and can optionally paste a job ad.
2. System generates an interview prompt.
3. User gets a thinking window, then a timed response window.
4. Frontend tracks posture and eye contact while recording audio.
5. Backend analyzes:
   • transcript quality and content,
   • vocal delivery characteristics,
   • posture/eye timeline data.
6. Results are displayed in a structured dashboard with strengths, improvement areas, and a next-step action plan.

• React 19 + Vite
• MediaPipe Tasks Vision
• Recharts
• CSS

• FastAPI + Uvicorn
• Groq/OpenAI-compatible SDK (for LLM + transcription)
• Librosa + NumPy + pydub (audio processing)
• ReportLab + Matplotlib (PDF/report chart generation)

• JavaScript / Python
• Git + GitHub

Interview-Bot/
├── backend/     # FastAPI API, prompt services, analysis pipeline
├── frontend/    # React app, interview flow, MediaPipe tracking, results UI
└── README.md    # Project overview (this file)

For folder-specific setup and architecture, see:

• frontend/README.md
• backend/README.md

cd backend
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -r requirements.txt
uvicorn app.main:app --reload

Backend default: http://127.0.0.1:8000

cd frontend
npm install
npm run dev

Frontend default: http://127.0.0.1:5173

• Open frontend URL.
• Select interview settings.
• Start interview and answer question.
• Review generated feedback on the results page.

• Keeping frontend and backend communication stable throughout the interview lifecycle.
• Uploading and processing recorded audio reliably.
• Tuning prompts so model output remained structured and useful.
• Iterating on CSS/UI to keep the experience clean and intuitive.
• Maintaining a consistent JSON response format across services.

• Built a fully functional end-to-end prototype under tight time constraints.
• Integrated live vision analysis, audio transcription, and AI feedback in one workflow.
• Produced a practical coaching tool that users can repeatedly train with.

• Better collaboration workflows with Git/GitHub.
• Practical audio ingestion and analysis for product use-cases.
• How to integrate LLM-driven feedback into a full-stack app.

Planned improvements include:

• More polished production UX.
• Accessibility enhancements (such as text-to-speech support).
• Historical tracking of interview sessions and progress over time.
• Stronger deployment/ops readiness for public usage.