🤖 AI Smart Interview Coach

Multi-Modal Biometric & Content Analysis System

An AI-powered coaching system that analyzes what you say, how you say it, and where you look.

View Live Demo · Report Bug

---

🚀 Overview

The Smart Interview Coach is an end-to-end evaluation platform designed to simulate high-stakes technical interviews. Unlike standard text-based tools, this system employs a multi-modal approach:

1. Computer Vision Pipeline: Tracks iris vectors in real-time to quantify eye contact confidence.
2. Audio Signal Processing: Analyzes pitch variance (Librosa) to detect monotone delivery.
3. Generative Reasoning: Uses Llama-3 (via Groq LPUs) to provide "Hiring Manager" feedback based on transcript context.

🏗️ System Architecture

graph TD
    A[User Video Input] --> B(Orchestrator)
    B -->|Visual Stream| C[Computer Vision Pipeline]
    B -->|Audio Stream| D[Audio Processing Pipeline]
    
    subgraph "Vision Stack"
    C --> C1[MediaPipe Face Mesh]
    C1 --> C2[Iris Landmark Extraction]
    C2 --> C3[Euclidean Vector Calculation]
    C3 --> C4[Gaze Ratio Score]
    end
    
    subgraph "Audio Stack"
    D --> D1[FFmpeg Extraction]
    D1 --> D2[Faster-Whisper ASR]
    D1 --> D3[Librosa Pitch Tracking]
    D3 --> D4[Tone Analysis Dynamic/Monotone]
    end
    
    D2 --> E[LLM Reasoning Engine]
    C4 --> E
    E -->|Context + Metrics| F[Llama-3.1 on Groq]
    F --> G[Final Feedback Report]

Loading