Loading...
Loading...
Works

Talent-match 2026

AI Resume Screener

Talent-Match is an AI-powered resume screener that reads between the lines. It doesn't just keyword-match "Python" and call it a day—it understands context.

💡 The Problem with Traditional ATS

Applicant Tracking Systems fail when a candidate says "built scalable microservices" instead of using the exact phrase "Spring Boot REST API". Humans get it. Old systems don't.

🤖 How It Works (Under the Hood)

  1. Document parsing: Extract text from PDFs/DOCX using Mammoth.js (preserves formatting context like bold headers)
  2. Semantic embedding: Convert resume + job description into vector embeddings using OpenAI's text-embedding-3-large model
  3. Cosine similarity: Calculate match score (0-100%) based on vector distance—"experienced in distributed systems" matches "built Kafka pipelines" even without shared keywords
  4. Explainability: GPT-4 generates human-readable reasoning: "Candidate's experience with Docker/K8s aligns with DevOps requirement"

✨ Smart Features

  • Skill inference: "Led team of 5 engineers" → automatically tags [Leadership, Team Management]
  • Experience weighting: Recent roles scored higher than 10-year-old internships
  • Red flag detection: Employment gaps, frequent job hopping, mismatched education—flagged for human review
  • Batch processing: Screen 100 resumes in 2 minutes vs. 2 hours manual review
🎯 Real-world impact: Reduced time-to-hire by 40% in pilot test with 500+ applicants for 3 engineering roles.

🔮 Technical Stack

Frontend React + TypeScript (drag-and-drop resume upload with progress bars)
Backend Node.js Express + OpenAI API (streaming responses for real-time feedback)
NLP Pipeline Custom prompt engineering with few-shot examples for consistency

Best use case: High-volume hiring for technical roles where skill depth matters more than keyword density

Talent-match
© 2026 Shorya Agarwal. All Rights Reserved.