🤖 AI-Powered Real-Time Mock Interview Coach

An end-to-end, multi-tier automated web platform designed to simulate realistic, adaptive technical interviews. The system captures live video and audio streams to perform real-time facial expression analysis, speech-to-text transcription, and natural language evaluation, providing detailed behavioral metrics and technical scoring breakdown graphs upon completion.

---

🏗️ Architectural Topology

The software ecosystem operates on a decoupled, three-tier architecture ensuring clean isolation of concerns:

 +-----------------------------------+
| Next.js Front-End | (Port 3000)
| React 19 / Tailwind CSS v4 |
+-----------------+-----------------+
|
REST / Audio | JSON Web Tokens
Video Blobs | (Auth Handshake)
v
+-----------------------------------+
| Node.js Express Server | (Port 5000)
| Authentication & Database |
+-----------------+-----------------+
|
Internal REST | Upstream Payload
Proxy Handlers | Forwarding
v
+-----------------------------------+
| FastAPI Backend | (Port 8000)
| PyTorch / OpenCV ML Inference |
+-----------------------------------+

1. Presentation Layer (/client) Built on Next.js 16 and React 19, managing high-frequency webcam visual loops (react-webcam) and recording audio tracks (react-media-recorder). Analytical trends are mapped with Recharts.

2. Orchestration Layer (/server) A reliable Express gateway driving structural storage tasks via Mongoose, validating state transitions, issuing JWT profiles, and managing multipart data pipelines via multer.

3. Machine Learning Layer (/backend) An asynchronous Python execution matrix fueled by FastAPI. It controls intensive CPU/GPU pipelines:

   • Text transcription (faster-whisper)
   • Text processing embeddings (sentence-transformers)
   • Visual computing (OpenCV)
   • Automated text feedback evaluation (language-tool-python)

---

📂 Project Structure Directory Matrix

├── backend/ # Python Asynchronous ML Pipeline
│ ├── app/
│ │ ├── knowledge_base/ # Curated technical prompt bases
│ │ ├── models/ # Pydantic schema validators
│ │ ├── modules/ # Core inferencing engines
│ │ │ ├── adaptive/
│ │ │ ├── evaluator/
│ │ │ ├── face_analysis/
│ │ │ ├── feedback/
│ │ │ ├── question_gen/
│ │ │ └── speech/
│ │ ├── routes/ # FastAPI routing paths
│ │ └── main.py # Python startup hub
│ └── requirements.txt
│
├── server/ # Node.js Express Session Tier
│ ├── src/
│ │ ├── config/
│ │ ├── middleware/
│ │ ├── models/
│ │ ├── routes/
│ │ └── index.js
│ └── package.json
│
└── client/ # Next.js Presentation App
├── src/
│ ├── app/
│ ├── components/
│ ├── context/
│ ├── hooks/
│ └── utils/
├── tailwind.config.js
└── package.json

---

⚡ Step-By-Step System Deployment

Prerequisites

Ensure your environment contains:

• Node.js: v18.x or above
• Python: v3.10.x or higher
• MongoDB: Local or cloud Atlas instance

---

Step 1: Initialize the Machine Learning Layer (/backend)

Navigate to backend

cd backend

Create virtual environment

python -m venv venv
source venv/bin/activate

Windows:

venv\Scripts\activate

Install dependencies

pip install -r requirements.txt

Download SpaCy model

python -m spacy download en_core_web_sm

Start FastAPI server

uvicorn app.main:app --host 127.0.0.1 --port 8000 --reload

---

Step 2: Initialize the Orchestration Gateway (/server)

Navigate to server

cd ../server

Install dependencies

npm install

Create .env

PORT=5000MONGO_URI=mongodb://127.0.0.1:27017/interview_coachJWT_SECRET=production_ready_cryptographic_randomized_hex_stringFASTAPI_URL=http://127.0.0.1:8000

Start server

npm run dev

---

Step 3: Initialize the Frontend Application (/client)

Navigate to client

cd ../client

Install dependencies

npm install

Create .env.local

NEXT_PUBLIC_API_URL=http://127.0.0.1:5000

Start frontend

npm run dev

Open:

http://localhost:3000

---

🔒 Environment Variables

Express Gateway (/server/.env)

Variable	Purpose	Example
PORT	Node server port	5000
MONGO_URI	MongoDB connection string	mongodb://127.0.0.1:27017/db
JWT_SECRET	JWT signing key	Secure random value
FASTAPI_URL	ML backend endpoint	http://127.0.0.1:8000

---

Next.js Client (/client/.env.local)

Variable	Purpose	Example
NEXT_PUBLIC_API_URL	Express backend URL	http://127.0.0.1:5000

---

📈 REST API Endpoint Registry

Authentication Routes

Method	Endpoint	Description
POST	/api/auth/register	Register new user
POST	/api/auth/login	Authenticate user

---

Session Control Routes

Method	Endpoint	Description
POST	/api/session/start	Start interview session
POST	/api/questions/next	Generate next question
POST	/api/evaluate/answer	Evaluate response
GET	/api/report/:sessionId	Fetch interview report

---

Python Backend Microservices

Method	Endpoint	Description
POST	/face/analyze	Facial expression analysis
POST	/speech/transcribe	Audio transcription
POST	/resume/extract	Resume parsing

---

🛠️ Troubleshooting

Camera / Microphone Access Issues

Ensure you are running on:

http://localhost

Modern browsers block media permissions on insecure origins.

---

MongoDB Connection Errors

Verify:

MongoDB Connected...

appears in the Node.js console logs.

---

PyTorch Performance Issues

The FastAPI backend automatically falls back to CPU inference if CUDA-enabled GPU drivers are unavailable.

---

🚀 Core Technologies

Frontend

• Next.js 16
• React 19
• Tailwind CSS v4
• Recharts
• Axios

Backend

• Node.js
• Express.js
• MongoDB
• Mongoose
• JWT
• Multer

AI / ML

• FastAPI
• PyTorch
• OpenCV
• Faster-Whisper
• Sentence Transformers
• SpaCy

---

📜 License

This project is intended for educational, research, and interview-preparation purposes.

---

👨‍💻 Author

Built as a scalable AI-assisted technical interview simulation platform using modern full-stack engineering and real-time machine learning inference pipelines.