HireGenius - AI Talent Intelligence Platform

Features

Inspiration

Traditional ATS systems reject 70% of qualified candidates for not matching exact keywords. A talented designer friend was rejected 47 times—not because she wasn't qualified, but because keyword matchers can't think like humans. We need AI that actually understands talent, not just matches words.

What it does

HireGenius analyzes candidates like an expert recruiter—holistically, not just keyword matching.

Core Features:

Skill Inflation Detection: Compares claimed skills vs. actual project evidence (catches resume padding)
Predictive Modeling: AI predicts ramp-up time, retention risk, and red flags
Batch Intelligence: Upload 50 resumes, get an auto-ranked leaderboard
Head-to-Head Duel: Compare any two candidates with AI explaining its reasoning
Interview Guides: Auto-generates role-specific questions with grading rubrics

The difference: We analyze evidence, not keywords. Did they actually lead teams or just attend meetings?

How we built it

Gemini 3 Integration:

2048-token thinking budget for multi-step reasoning (enables retention risk predictions)
Structured JSON output with enforced schemas (prevents UI-breaking responses)
Expert prompt engineering (AI acts as strict recruiter requiring evidence)

Tech Stack: React 19 + TypeScript, Tailwind CSS, Recharts, JSZip Architecture: 100% client-side (no backend—all AI inference in browser)

Challenges we ran into

AI Hallucination Prevention: Early versions confidently invented facts. Solution: aggressive prompt engineering requiring evidence citations + confidence scoring that drops when AI infers vs. cites.

Batch Processing Speed: 50 resumes would take 5+ minutes sequentially. Implemented concurrent API calls (5 at a time) with real-time progress UI and optimistic updates.

Accomplishments that we're proud of

Skill Inflation Detection — Compares claimed skills against actual usage in work history. Could save companies thousands in bad hires.

Transparent AI Reasoning — Head-to-Head Duel mode shows why the AI chose Candidate A over B (e.g., "deeper backend experience at Stripe").

Gemini 3's Thinking Budget in Action — Visible predictions requiring multi-step logical chains (retention risk, cultural fit).

What we learned

Structured Output is Production-Critical: Before enforcing JSON schemas, 20% of responses broke the UI. After: 0%.

AI Transparency Builds Trust: Showing why the AI picked someone matters more than just showing who won.

Domain Expertise > General AI: Encoding recruiting best practices into prompts makes the AI perform like a specialist, not a chatbot.

What's next for HireGenius - AI Talent Intelligence Platform

Short-term: Multi-language support, resume anonymization (reduce bias), candidate feedback loops

Vision: Eliminate the resume entirely. Paste a job description → AI automatically sources, screens, and ranks the top 10 candidates from a global talent pool with full explainability. Gemini 3's reasoning makes this possible for the first time.

Built with Gemini 3 Pro for the Gemini 3 Global Hackathon