Hiro.ai
A recruiter before recruiters ever see your application.
Hiro AI (Target the Role) is an agentic ATS optimization platform that bridges the semantic and structural gap between an applicant's resume and the job description. It runs as a multi-agent graph — an ATS scanner, a senior resume writer, and a QA auditor working in unison to optimize a profile with lexical and mathematical precision.
Existing 'AI resume' tools are primitive keyword stuffers. They hallucinate skills, ignore ATS structure, and give applicants no signal on how a real recruiter would score their fit.
Live at gethiroaiapp.tech with precise keyword alignment, automated STAR-structured bullet rewrites, and 99.9%+ availability from the dual-layer failover — reasoning surfaced instead of hidden.
How it was built.
- 01
Multi-Agent Feedback Loop: a LangGraph-inspired state graph with an ATS Analyzer (semantic vector distance + keyword density), a Lead Resume Rewriter (STAR-methodology bullets, no hallucinated experience), and a Self-Reflection QA Auditor that triggers re-iteration when compliance criteria fail.
- 02
Hybrid Resiliency Architecture: primary route flows through a Node.js/Express gateway into the Python orchestration engine; on latency or disconnects the frontend seamlessly falls back to a browser-side Gemini SDK sandbox — zero-downtime UX.
- 03
Production-grade auth: Supabase authentication with JWT injection into all API request headers (Authorization: Bearer <token>) — replacing demo/mock auth wrappers.
- 04
Stack: React 18 + Vite + TypeScript + Tailwind on the client, @google/genai (Gemini 3.5 / 2.5 Flash) for reasoning, Node.js/Express gateway, Python LangGraph-style workflows, Supabase for auth + persistence.
"The hardest part was making agents disagree productively — the QA Auditor only ships a rewrite when the ATS Analyzer and Resume Rewriter reach compliance quorum."