Inspiration
Fresh graduates and early-career developers apply to dozens of jobs with the same generic CV. Most applications are filtered out by keyword screening before a person reads them. Tailoring each CV by hand takes about an hour per job. Generic AI rewriters are fast, but they often invent skills and numbers, which can cost you the job at the interview.
What it does
You paste a job post and your CV. HireReady runs an agentic workflow and shows every step live:
- Parses the job (must-haves, ATS keywords) and your CV into structured data.
- Analyzes gaps: gives an original match score, matched skills, missing skills with fast ways to close them, and transferable strengths you didn't highlight.
- Drafts a tailored CV that never invents facts. Missing metrics become
[X%]placeholders for you to fill in. - Self-critiques: a separate critic agent scores the draft as an ATS screener and fact-checks every claim against your original CV. The writer revises until the score is at least 85 and no claim is flagged (max 2 revisions).
- Prepares you: writes a targeted cover letter and predicts 6 interview questions, including ones about your gaps, with STAR answer outlines.
Output in English, Vietnamese or Bangla. You can copy, download or print to PDF.
How we built it
- Next.js 15 + TypeScript, deployed on Vercel.
- The agent workflow (
lib/agent.ts) mixes parallel steps, structured JSON outputs and a writer ↔ critic feedback loop with a stopping condition. - Agent events are streamed to the UI as NDJSON, so users see each step, its timing and its result as it happens.
- Meta Model API (
muse-spark-1.3) powers the agents; the client works with any OpenAI-compatible provider (OpenRouter fallback), so models can be swapped with env vars. - Devin was used during development to (fill in: e.g. scaffold the project, implement the streaming endpoint, write the UI, fix bugs).
Challenges
- Stopping the model from "improving" a CV by inventing experience. We solved it with a dedicated fact-checking critic, which beats prompt instructions alone.
- Getting reliable structured output across different models (JSON mode + a repair fallback).
What's next
- PDF/DOCX CV upload
- Track applications across many jobs
- Browser extension: tailor straight from a LinkedIn or BDJobs job page
Target users
Students, fresh graduates and early-career professionals in Bangladesh and beyond who apply to many jobs and need each application to stand out.
Try it
- Live demo: https://hireready-agent.vercel.app/
- Repo: https://github.com/AIVIETNAM-AIO-felixdoit/hireready-agent
- Demo video: https://youtu.be/LZQsHbsydGY
- Click Load sample, then Run agent. A full run takes about 1–1.5 minutes.
Built With
- meta-model-api
- nextjs
- react
- typescript
- vercel