🚀 Career Connect AI

AI-Powered Recruitment Automation Platform

Semantic Resume Screening · Skill Gap Analysis · AI Interviews · Multimodal Evaluation

---

📌 Overview

Career Connect AI is a full-stack recruitment automation platform that replaces manual resume screening with an AI-driven pipeline. It uses Doc2Vec embeddings, GPT-4.1 semantic reasoning, and upcoming multimodal analysis (emotion, speech, identity) to give HR teams objective, data-backed hiring decisions — and gives candidates a personalised roadmap to close their skill gaps.

---

✨ Features

✅ Implemented (Phase 1 & 2)

Module	Feature
🔐 Auth	JWT-based register/login, role-based access (Candidate / HR / Admin)
📄 Resume Upload	PDF & DOCX parsing, Doc2Vec embedding on upload
💼 Job Descriptions	HR creates JDs with title, description & required skills; auto-embedded
🧠 Semantic Matching	Doc2Vec cosine similarity + GPT-4.1 hybrid score (FR-4.1, 4.4)
🎨 Match Score UI	Circular progress rings, colour-coded tier (Green / Amber / Red) (FR-4.2)
🔍 Skill Overlap	Matched, missing, and extra skills (FR-4.3)
📊 Recruiter View	Sortable and filterable ranked candidate table (FR-4.5)
🗺️ Skill Gap Analysis	AI-detected missing skills with impact ranking (FR-5.1, 5.3)
📚 Learning Recommendations	GPT-4.1 course names, topic descriptions, estimated time (FR-5.2)
📖 Smart Article Redirect	Start Learning opens best GPT-chosen tutorial in new tab
✅ Progress Tracking	Mark skills In Progress / Completed; full event history (FR-5.4)
🤖 AI Interview	Dynamic JD-contextualised question generation via GPT-4.1 RAG

🔜 Roadmap (Phase 3+)

Module	Status
😐 Emotion Analysis	DeepFace integration during interview — planned
🔊 Speech Analysis	Librosa + RAVDESS tone/sentiment — planned
🛡️ Anti-Cheat Engine	YOLOv8 multi-person detection + tab-switch guards — planned
🪪 Identity Verification	Face-match against uploaded ID — planned
📄 PDF Reports	ReportLab evaluation reports — planned
📨 Event-Driven Pipeline	RabbitMQ / Kafka for async processing — planned
🐳 Containerisation	Docker + Kubernetes deployment — planned

---

🏗️ Architecture

graph TB
    subgraph FE["Frontend - React + Vite + Tailwind"]
        UI_Auth[Auth Pages]
        UI_Dashboard[Dashboard]
        UI_Match[Resume Match]
        UI_Interview[AI Interview]
        UI_Reports[Reports]
    end

    subgraph BE["Backend - FastAPI"]
        API[REST API port 8000]

        subgraph Routers["Routers"]
            R_Auth["auth"]
            R_Resume["resume"]
            R_JD["jd"]
            R_Match["match"]
            R_Interview["interview"]
            R_Rec["recommendations"]
            R_HR["hr"]
        end

        subgraph AIServices["AI Services"]
            SVC_AI["ai_service.py - GPT-4.1"]
            SVC_D2V["doc2vec_service.py"]
            SVC_GEM["gemini_service.py shim"]
        end

        subgraph CoreLayer["Core"]
            SEC["security.py - JWT"]
            CFG["config.py - Settings"]
            MODELS["models.py - ORM"]
        end
    end

    subgraph Storage["Storage"]
        DB[(SQLite or PostgreSQL)]
        FS[Doc2Vec Artifacts]
    end

    subgraph ExtAI["External AI"]
        GHAI["GitHub AI - GPT-4.1"]
    end

    FE -->|HTTP + JWT| API
    API --> Routers
    Routers --> AIServices
    Routers --> CoreLayer
    AIServices --> GHAI
    AIServices --> SVC_D2V
    CoreLayer --> MODELS
    MODELS --> DB
    SVC_D2V --> FS

Loading

---

🔄 Resume Matching Pipeline

sequenceDiagram
    participant C as Candidate
    participant API as FastAPI
    participant D2V as Doc2Vec
    participant GPT as GPT-4.1
    participant DB as Database

    C->>API: POST resume upload PDF or DOCX
    API->>D2V: Extract text and infer embedding
    D2V-->>API: 100-dim vector
    API->>DB: Store resume and embedding

    C->>API: POST match with resumeID and jobID
    API->>DB: Fetch both embeddings
    API->>API: Compute cosine similarity
    API->>GPT: Semantic reasoning prompt
    GPT-->>API: score 0 to 100
    API->>API: hybrid = cosine 70pct + GPT 30pct
    API->>GPT: ai_skill_recommendations for missing skills
    GPT-->>API: courses and descriptions
    API->>DB: Persist SkillRecommendation rows
    API-->>C: hybridScore + tier + skillOverlap + recommendations

Loading

---

🤖 AI Interview Flow

sequenceDiagram
    participant C as Candidate
    participant API as FastAPI
    participant GPT as GPT-4.1
    participant DB as Database

    C->>API: POST interview start with jobID and experience
    API->>GPT: ai_generate_question with JD context
    GPT-->>API: contextual first question
    API->>DB: Store session and transcript
    API-->>C: sessionID and firstQuestion

    loop Each answer turn
        C->>API: POST interview answer with transcript
        API->>GPT: ai_generate_question with history
        GPT-->>API: next contextual question
        API-->>C: nextQuestion
    end

    C->>API: POST interview end
    API->>API: create evaluation for session
    API-->>C: evalID and estimatedReady

Loading

---

🗂️ Project Structure

Career-Connect-AI/
├── src/
│   ├── pages/
│   │   ├── ResumeMatch.tsx         # Full matching and skill gap UI
│   │   ├── InterviewSelection.tsx
│   │   ├── Dashboard.tsx
│   │   ├── Reports.tsx
│   │   └── Profile.tsx
│   ├── context/AuthContext.tsx
│   ├── lib/api.ts
│   └── index.css
│
└── backend/
    ├── app/
    │   ├── main.py
    │   ├── models.py
    │   ├── schemas.py
    │   ├── db.py
    │   ├── deps.py
    │   ├── security.py
    │   ├── ai_service.py           # GPT-4.1 unified service
    │   ├── doc2vec_service.py
    │   ├── gemini_service.py
    │   ├── utils.py
    │   ├── core/config.py
    │   ├── artifacts/
    │   └── routers/
    │       ├── auth.py
    │       ├── resume.py
    │       ├── jd.py
    │       ├── match.py
    │       ├── interview.py
    │       ├── evaluation.py
    │       ├── recommendations.py
    │       ├── report.py
    │       ├── dashboard.py
    │       ├── history.py
    │       └── profile.py
    ├── requirements.txt
    └── .env

---

🛠️ Tech Stack

Layer	Technology
Frontend	React 18, Vite, TypeScript, Tailwind CSS, Lucide Icons
Backend	FastAPI, Uvicorn, Python 3.11+
Database	SQLite (dev) / PostgreSQL (prod) via SQLAlchemy
AI — Primary	GPT-4.1 via GitHub AI Inference
AI — Embeddings	Doc2Vec (Gensim)
AI — Fallback	TF-IDF cosine similarity
Auth	JWT (python-jose) + bcrypt (passlib)
File Parsing	PyPDF2, python-docx

---

⚡ Quick Start

Prerequisites

• Python 3.11+, Node.js 18+
• A GitHub PAT with model access

1. Clone

git clone https://github.com/Shreyyy07/Career-Connect-AI---Major-Project1.git
cd Career-Connect-AI---Major-Project1

2. Backend

cd backend
python -m venv venv
venv\Scripts\activate          # Windows
pip install -r requirements.txt

Create backend/.env:

DATABASE_URL=sqlite+pysqlite:///./career_connect_ai.db
JWT_SECRET=your-secret-key-here
CORS_ORIGINS=http://localhost:5173
GITHUB_TOKEN=github_pat_xxxxxxxxxxxx
GITHUB_AI_ENDPOINT=https://models.github.ai/inference
GITHUB_AI_MODEL=openai/gpt-4.1

python -m uvicorn app.main:app --reload
# API: http://localhost:8000
# Swagger: http://localhost:8000/docs

3. Frontend

cd ..
npm install
npm run dev
# App: http://localhost:5173

---

📡 Key API Endpoints

Method	Endpoint	Description
POST	/api/v1/auth/register	Register user
POST	/api/v1/auth/login	Login and get JWT
POST	/api/v1/resume/upload	Upload PDF or DOCX
POST	/api/v1/jd/upload	Create JD (HR only)
POST	/api/v1/match	Run hybrid AI match
GET	/api/v1/hr/matches	Recruiter ranked view
GET	/api/v1/recommendations/{id}/resource-url	GPT article URL
POST	/api/v1/recommendations/{id}/status	Update learning status
POST	/api/v1/interview/start	Start AI interview
POST	/api/v1/interview/answer	Submit answer and get next question
POST	/api/v1/interview/end	End session and trigger evaluation
GET	/api/v1/health	Health check

---

🧠 AI Service Functions

All in backend/app/ai_service.py with graceful offline fallbacks:

Function	Purpose	Fallback
ai_semantic_score()	Resume–JD match score 0 to 100	TF-IDF cosine
ai_generate_question()	Dynamic interview question from JD	Static question bank
ai_evaluate_answer()	Score answer quality	50.0 default
ai_skill_recommendations()	Courses, descriptions, time estimates	Empty list
ai_find_resource_url()	Best tutorial URL for a skill	Google search URL

---

🗃️ Database Schema

User ──────────┬───── Resume (embedding_csv)
               ├───── JobDescription (embedding_csv, skills_csv)
               ├───── InterviewSession ── Evaluation
               ├───── Assessment
               └───── SkillRecommendation ── SkillRecommendationEvent

---

🔒 Security

• Bearer JWT required on all endpoints except /auth/*
• Passwords hashed with bcrypt
• Role-based access: candidate | hr | admin
• CORS restricted to configured origins

---

📄 License

MIT — see LICENSE for details.

---

Built with ❤️ by Shreyyy

⭐ Star this repo if you find it useful!