📄
1,000+
Bulk Intake Capacity
🎯
3
Keka Engineering JDs
< 1.9s
1,000 Resumes Bulk Scan
🛡️
3 Layers
Deduplication Protection
⚙️ Screening Configuration
Instant Analysis
— OR UPLOAD CUSTOM RESUME FILE —

No Candidate Evaluated Yet

Select a candidate on the left or upload a file and click "Analyze & Score Candidate" to generate instant fit scoring, skills breakdown, and deduplication verification.

⚡ Keka Bulk Screening Engine — 1,000+ Applications Intake
High Throughput Celery / Async Pool

When processing thousands of job applications across client organizations every day, sequential API calls bottleneck recruitment. Our high-concurrency bulk screening engine ingests 1,000 applicant resumes simultaneously, verifies exact & semantic duplicates, evaluates multi-category fit scores, and shortlists the top 50 candidates in under 2 seconds.

📊 Keka ATS Batch Intelligence Leaderboard (10 Test Resumes)

Simulates Keka's automated ATS pipeline processing 10 applicant resumes simultaneously against the Senior Python Backend Engineer Job Description. Automatically calculates fit scores, ranks applicants from highest to lowest fit, and highlights exact & semantic duplicates.

Batch Pipeline Ready

Click the "Run 10-Candidate Batch Screening" button above to evaluate all sample applicants instantly.

🛡️ Candidate Deduplication & Anti-Fraud Engine
Active Protection

When processing thousands of job applications across client organizations every day, Keka's ATS must catch candidates who resubmit identical resumes, submit slightly reformatted CVs with the same contact details, or get uploaded by multiple staffing agencies. Our POC features a 3-layer deduplication engine:

1. Exact SHA-256 Hash

Computes a cryptographic SHA-256 fingerprint on normalized text. Catching 100% identical file uploads instantly with zero overhead.

2. Semantic Profile Key

Extracts and normalizes candidate email (`arjun.verma.py@gmail.com`) and phone (`+919876543210`). Flags duplicates even when candidate changes resume template.

3. Fuzzy Content Similarity

Uses SequenceMatcher / Jaccard similarity to detect high content overlap (>88%), identifying paraphrased or updated resumes submitted by the same candidate.

Test Deduplication with Live Sample Resumes

Test File Candidate Name Email / Phone Expected Status Action
resume_1_arjun_verma_sr_python.pdf Arjun Verma arjun.verma.py@gmail.com / +919876543210 Original Submission
resume_3_arjun_verma_duplicate.pdf Arjun Verma (Reformatted CV) arjun.verma.py@gmail.com / +919876543210 ⚠️ Semantic Profile Match
resume_8_rohit_kumar_fullstack.pdf Rohit Kumar rohit.kumar.fs@gmail.com / +919667788990 Original Submission
resume_9_rohit_kumar_duplicate.pdf Rohit Kumar (Duplicate Layout) rohit.kumar.fs@gmail.com / +919667788990 🔴 High Fuzzy Similarity
📈 Google Sheets & Excel ATS Audit Log
📥 Download Full ATS Report (.xlsx)

Every candidate evaluated through the single or batch scoring endpoints is automatically recorded in our real-time audit logger. Supports direct synchronization to Google Sheets via Service Account API or fallback local CSV/Excel generation.

Timestamp Candidate Name Experience Location Job Description Fit Score Recommendation Is Duplicate Scoring Engine
Loading logged evaluations...
🏗️ Keka Resume Intelligence — Full Stack System Architecture
SaaS Production Ready

Frontend Architecture (`frontend/`)

Our frontend is engineered as a zero-latency Single Page Application (SPA) designed with enterprise dark-mode aesthetics, responsive CSS grids, and dual-mode connectivity:

  • Live Backend Sync Mode: Communicates with `FastAPI` via REST APIs (`/api/score-sample`, `/api/bulk-scan-1000`, `/api/batch-score`) for real-time PDF parsing and bulk ingestion.
  • In-App Preview Mode: Automatically detects when viewed inside sandboxed file previewers (`present_file`) and seamlessly runs full client-side interactive simulations for 1,000+ resumes.
  • Interactive Data Visualizations: Renders fit score breakdown bars, recommendation distributions, and multi-tier deduplication alerts.

Backend Architecture (`backend/ & main.py`)

Our backend is powered by **FastAPI** (`Python 3.13`), **pypdf**, and **Groq LLaMA 3.3 70B** to process high-throughput applicant data with extreme reliability:

  • `bulk_generator.py`: High-concurrency 1,000+ resumes batch simulation & scoring engine with embedded deduplication signatures.
  • `parser.py`: Extracts raw text from PDFs/DOCX and computes exact SHA-256 + semantic profile fingerprints.
  • `scorer.py`: Routes scoring prompts to Groq LLaMA or OpenAI API, with deterministic local heuristics fallback.
  • `sheets.py`: Synchronizes scores to Google Sheets (`gspread`) and exports `keka_ats_scoring_report.xlsx` and `keka_1000_resumes_bulk_report.xlsx`.

Project Repository Directory Structure

keka-resume-intelligence/
├── main.py                 # FastAPI application & REST API endpoints
├── scorer.py               # Groq LLaMA / OpenAI / Local heuristic scoring engine
├── parser.py               # Multi-format document parser & 3-layer deduplication
├── bulk_generator.py       # 🚀 1,000+ Resumes Bulk Screening & Scoring Engine
├── sheets.py               # Google Sheets logger & Excel report builder
├── requirements.txt        # Backend Python dependencies
├── .env                    # Environment configuration & API keys
├── frontend/               # Dedicated Frontend Application
│   ├── index.html          # Enterprise SPA Web Dashboard (Interactive UI)
│   ├── style.css           # Responsive Dark-Mode UI stylesheet
│   └── app.js              # Client-side API client & interactive lab controller
├── sample_resumes/         # 10 realistic PDF test resumes across varied roles & fit levels
├── sample_jd/              # Keka Engineering Job Descriptions (Backend, AI/ML, Fullstack)
├── keka_ats_scoring_report.xlsx # Automatically generated Excel report (10 resumes)
└── keka_1000_resumes_bulk_report.xlsx # Complete 1,000-candidate Bulk Screening Excel report