Building Resumind AI Resume Analyzer

The problem

Applicant tracking systems reject resumes for formatting and keyword gaps that a candidate cannot see. Kush Gangwal built Resumind so a job seeker can upload a resume, describe a target role, and get a structured report instead of vague advice.

How Resumind works

Resumind parses the uploaded document, keeps a structured representation of sections and skills, and sends that representation to an LLM with a tight rubric: ATS readability, keyword coverage, impact language, and missing evidence. The model returns JSON, not a blob of marketing copy, so the UI can show scores and rewrite tips side by side.

• Resume upload and section extraction
• Role-aware ATS scoring
• Category tips for skills, formatting, and keywords
• Saved reports so a candidate can compare versions

What Kush Gangwal learned

AI-powered applications fail when the model is asked to 'be helpful' with no schema. Resumind treats the model as a structured reviewer. That is the same approach Kush Gangwal uses when integrating AI APIs in production work at SK Groups and on Potato Bazaar.

Building Resumind AI Resume Analyzer also forced clear UX: people need a score they can trust and a next edit they can make in ten minutes.

Links

• GitHub

More projects