Interviewmate

Voice-first AI mock interviews that feel real, with actionable feedback to improve technical + communication performance.

Problem

• Low realism → high anxiety: candidates underperform in real interviews.
• Coaching is expensive: most people can’t access quality prep.
• Generic feedback: doesn’t reveal exact weak points.
• Peer mocks are inconsistent: low rigor, low objectivity.

Solution (Interviewmate)

• Human-like interviews using Generative AI + real-time voice.
• Deep follow-ups powered by Google Gemini.
• Ultra-low latency voice via Deepgram (STT) + ElevenLabs (TTS).
• Performance analytics: confidence, clarity, structure, technical depth.
• Multiple tracks: HR, Technical, Managerial — role/company aligned.

Target Users

• Job seekers targeting top tech roles.
• Career switchers validating skills in realistic scenarios.
• Students preparing for placements.
• Hiring teams using it for candidate readiness practice.

Guarantees

• Privacy-first: sessions are encrypted and secure.
• Zero-lag experience: websocket-based real-time architecture.
• Industry-aligned questions: generated per role and current trends.

Architecture

High-Level System Design

graph TD
    User([User]) <-->|Websocket/MediaStream| Client[React Frontend]
    Client <-->|Socket.io/REST| Server[Node.js Backend]
    
    subgraph "AI & Voice Cloud"
        Server <-->|Gemini API| Gemini[Reasoning & Evaluation]
        Server <-->|Deepgram SDK| STT[Speech-to-Text]
        Server <-->|ElevenLabs SDK| TTS[Text-to-Speech]
    end
    
    subgraph "Data & Auth"
        Server --- DB[(MongoDB)]
        Server --- Auth[Google OAuth]
        Server --- Pay[Razorpay]
    end

Loading

End-to-End Process Flow

sequenceDiagram
    participant U as User
    participant C as Client (React)
    participant S as Server (Node.js)
    participant AI as Gemini / ElevenLabs
    
    U->>C: Start Interview
    C->>S: Init Session (Websocket)
    S->>AI: Generate First Question
    AI-->>S: Text Content
    S->>AI: Text-to-Speech
    AI-->>S: Audio Stream
    S->>C: Push Audio + Text
    C->>U: Play Audio
    U->>C: Speak Response (Audio)
    C->>S: Stream Audio
    S->>AI: Speech-to-Text (Deepgram)
    AI-->>S: Validated Transcript
    S->>AI: Process & Evaluate
    S->>U: ... repeat for all questions ...
    S->>C: Generate Final PDF Report

Loading

<<<<<<< HEAD

ER Diagram

erDiagram
    USER ||--o{ INTERVIEW : "starts"
    USER ||--o{ PAYMENT : "makes"
    USER ||--o{ TOKEN_TRANSACTION : "uses"
    INTERVIEW ||--|| REPORT : "generates"
    INTERVIEW ||--o{ SESSION_CHUNK : "contains"
    
    USER {
        string name
        string email
        string subscription
        int tokenBalance
    }
    INTERVIEW {
        string role
        string type
        int overallScore
        string status
    }
    REPORT {
        json skillBreakdown
        string feedback
    }

Loading

Tech Stack

Layer	Technologies
Frontend	React, Tailwind CSS, Framer Motion, Socket.io-client, Radix UI, Recharts
Backend	Node.js, Express.js, Socket.io, Mongoose
AI (LLM)	Google Gemini Pro
Voice (STT/TTS)	Deepgram, ElevenLabs
Database	MongoDB Atlas
Cloud/Infrastructure	Razorpay, Google OAuth 2.0, jsPDF

Project Structure

interviewmate/
├── client/                 # Frontend - React (Vite)
│   ├── src/
│   │   ├── components/     # UI Components (Radix, Framer)
│   │   ├── pages/          # Interview, Dashboard, Landing
│   │   ├── services/       # API & Socket handlers
│   │   └── context/        # Auth & State management
├── server/                 # Backend - Node.js
│   ├── models/             # Mongoose Schemas (User, Interview, etc.)
│   ├── routes/             # REST Endpoints
│   ├── services/           # AI, STT, TTS logic
│   ├── websocket/          # Real-time interview handlers
│   └── middleware/         # Auth & Validation
├── .env                    # Environment variables
└── package.json            # Dependencies & Scripts

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Built with ❤️ for candidates everywhere.