# Kumud Waykole

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [KUMUD WAYKOLE](https://cerebralvalley.ai/u/kumudwaykole)
- **GitHub:** https://github.com/kumudwaykole/Ai-resume-builder
- **Demo video:** https://drive.google.com/drive/folders/1DPHkgAUt3vKcPJ1l6sXneWCi2494wWuF?usp=sharing
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5326

SkillBridge AI helps job seekers show what they can actually do, and helps hiring teams see it.

The problem: candidates rewrite the same resume for every posting without knowing what a role really requires, and many AI resume tools "improve" resumes by inventing achievements. Recruiters then receive polished resumes they can't verify.

SkillBridge separates facts from presentation. Each candidate keeps one fact sheet (roles, dates, highlights, skills, education, certifications). AI output is always checked against those facts, so the platform never makes things up.

What works today:
• Job analysis: paste any posting and a Meta model turns it into a structured spec: must-haves vs nice-to-haves, required years, canonical skills, keywords to cover and likely interview topics. It runs in a background queue with live progress streamed to the browser, and an identical posting is served instantly from cache.
• Profile (fact sheet) with a live strength meter, validation and conflict-safe saving across tabs and devices.
• Resume import backend: an uploaded PDF or DOCX is checked by its bytes, text-extracted in a worker, structured by the model with a source quote for every item, and merged into the profile only after the candidate confirms it.
• Secure accounts with rotating refresh tokens and CSRF protection, candidate and recruiter roles, light and dark themes, and phone layouts.

Next on the roadmap: evidence-linked matching of a profile against a job, a personal gap plan, and an adaptive voice interviewer built on Meta's realtime transcription model. It asks follow-ups based on what the candidate actually said and produces a report with verified quotes and no single score, so people, not the AI, make hiring decisions.

Built solo with NestJS, Next.js 16, PostgreSQL, Redis and BullMQ, Zod and Tailwind CSS.

## More from Meta Global AI Developer Hackathon

- [Videonerd](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5323)
- [Proof Loop](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5324)
- [Sashi Perera](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5325)
- [zorith](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5327)
- [prompt2brand](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5328)
- [AegisLabel-FX](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5329)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5326. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt