CogniHire — AI Based Resume Analyser

Upload your resume. Paste a job description. Get your ATS score, skill gaps, rewritten bullet points, interview prep, and matching job listings — all in one place, with your progress tracked over time.

---

What It Does

Most resume tools do keyword matching. CogniHire goes further:

• ATS Scoring — Hybrid score combining keyword match + semantic similarity (sentence-transformers)
• Skill Gap Analysis — Custom-trained spaCy NER model extracts skills from both resume and JD, identifies what's missing
• AI Feedback — Groq (Llama 3.3 70B) gives actionable recruiter-style feedback
• Resume Rewrites — Weak bullet points selectively rewritten (only where genuinely needed) to better highlight real, already-listed skills — never fabricated metrics or invented achievements
• Interview Prep — 10 role-specific questions (technical, conceptual, behavioral) with one-line answering tips
• JD Scraping — Paste a LinkedIn/Naukri/Indeed URL instead of copying the job description manually
• AI-Predicted Job Matching — Groq predicts the 5 job roles best suited to a resume, then pulls live listings for each via the JSearch API
• Score History & Trend Tracking — Every analysis is saved to Firestore; a dashboard chart shows ATS score progress across resume versions over time
• PDF Report Export — Download the full analysis (scores, skills, rewrites, interview prep) as a branded PDF report
• Authentication — Email/password and Google OAuth sign-in via Firebase, with analyses tied to your account

---

Screenshots

---

Tech Stack

Layer	Technology
Frontend	React 19, Vite, Tailwind CSS
Backend	FastAPI, Python 3.12
NLP	Custom spaCy NER model + EntityRuler + skill taxonomy
Embeddings	all-MiniLM-L6-v2 via sentence-transformers
LLM	Llama 3.3 70B via Groq API
PDF Parsing	PyPDF2 + Tesseract OCR (fallback for scanned PDFs)
JD Scraping	BeautifulSoup + Requests
Job Matching	JSearch API (RapidAPI) — aggregates LinkedIn, Indeed, Naukri
Auth & Storage	Firebase Authentication + Firestore
Report Export	jsPDF (client-side PDF generation)

---

Architecture

Resume + JD  →  FastAPI Backend  →  React Frontend  →  Firebase
                 │                    │
   parser.py ─── PDF/OCR extraction   Sidebar: Dashboard · Skills ·
   scorer.py ─── spaCy NER + SBERT    Rewrites · Interview · Jobs ·
   advisor.py ── Groq (feedback,      History
                  rewrites, roles)    PDF export · Auth · Dark mode
   scraper.py ── JD scraping
   job_search.py ─ JSearch API

Full pipeline: resume PDF → text extraction (with OCR fallback) → spaCy NER skill extraction → hybrid ATS scoring (keyword + semantic) → three Groq calls (feedback, rewrites, interview prep) → results rendered across a sidebar-based dashboard, saved to Firestore for history tracking.

---

Local Setup

Prerequisites

• Python 3.10+
• Node.js 18+
• Tesseract OCR installed (Windows / Mac)
• Groq API key (free at console.groq.com)
• RapidAPI account with JSearch API subscribed (free tier at rapidapi.com)
• Firebase project with Authentication (Email/Password + Google) and Firestore enabled (console.firebase.google.com)

Backend

cd backend
python -m venv venv

# Windows
venv\Scripts\activate
# Mac/Linux
source venv/bin/activate

pip install -r requirements.txt

Create a .env file inside backend/:

GROQ_API_KEY=your_groq_api_key_here
JSEARCH_API_KEY=your_rapidapi_key_here

Start the server:

uvicorn main:app --reload

Backend runs at http://localhost:8000 API docs at http://localhost:8000/docs

Frontend

cd frontend
npm install
npm run dev

Create a .env file inside frontend/:

VITE_FIREBASE_API_KEY=your_api_key
VITE_FIREBASE_AUTH_DOMAIN=your_auth_domain
VITE_FIREBASE_PROJECT_ID=your_project_id
VITE_FIREBASE_STORAGE_BUCKET=your_storage_bucket
VITE_FIREBASE_MESSAGING_SENDER_ID=your_messaging_sender_id
VITE_FIREBASE_APP_ID=your_app_id

Frontend runs at http://localhost:5173

---

Project Structure

CogniHire/
├── backend/          # FastAPI app, spaCy NER model, Groq/scraping/job-search logic
├── frontend/          # React app — components, pages, auth, PDF export
├── screenshots/
└── README.md

---

API Reference

POST /analyze

Field	Type	Description
resume	File	PDF resume (required)
jd_text	Form	Job description as plain text
jd_url	Form	Job posting URL (LinkedIn, Naukri, Indeed)

Response:

{
  "score": {
    "ats_score": 79.54,
    "keyword_score": 87.5,
    "semantic_score": 71.59,
    "matched_skills": ["react", "nodejs", "python"],
    "missing_skills": ["mongodb", "docker"]
  },
  "advice": "...",
  "rewrites": "...",
  "interview_questions": "...",
  "resume_text": "..."
}

POST /jobs

Field	Type	Description
resume	File	PDF resume (required)
location	Form	Job search location (default: "India")

Response:

{
  "predicted_roles": ["Full Stack Developer", "React Developer", "..."],
  "jobs": [
    {
      "title": "Senior React Developer",
      "company": "Example Corp",
      "location": "Bangalore",
      "employment_type": "FULL_TIME",
      "apply_url": "https://...",
      "matched_role": "React Developer",
      "description": "...",
      "required_skills": ["..."]
    }
  ]
}

---

What Makes This Different

Most resume tools check if keywords match. CogniHire tries to go a step further — understanding skill aliases, scoring on meaning rather than exact words, and rewriting only what genuinely needs fixing without inventing anything new. The idea was to build something that actually helps during a real job search, not just something that produces a score.

Author

Pranali Pathak GitHub

---

Built from scratch — from a Google Colab notebook to a full-stack production-ready web app with auth, persistent history, and AI-powered job matching.