Lumina

Lumina is an AI-powered talent intelligence platform that transforms raw resumes and GitHub portfolios into recruiter-ready, narrative-rich and verified candidate profiles — instantly, in bulk.

Comment

Lumina - AI-Powered Talent Intelligence Platform

Inspiration

Recruiters spend countless hours manually screening resumes, cross-referencing portfolios, and trying to verify candidate claims. We built Lumina to automate this tedious process and deliver recruiter-ready intelligence the moment a resume enters the pipeline.

What it does

Lumina is an AI-powered recruiting platform that transforms raw resumes and candidate portfolios into comprehensive, narrative-rich candidate profiles. The platform:

  • Bulk Resume Processing - Upload multiple resumes at once for a job position
  • Intelligent Parsing - Extracts structured data from PDFs including skills, experience, and education
  • Smart Candidate Ranking - Uses semantic matching with Pinecone vector search to score candidates against job requirements
  • Deep Portfolio Analysis - Automatically analyzes GitHub profiles through a 4-phase agentic workflow: a. Context Gathering - Extracts candidate technical profile from resume b. Repository Discovery - Intelligently filters and ranks repositories by relevance c. Deep Analysis - Evaluates code quality, validates resume claims, identifies strengths/concerns d. Final Synthesis - Generates comprehensive hiring assessment with recommendations
  • Resume Validation - Cross-references GitHub projects against resume claims to verify authenticity
  • Comprehensive Scoring - Provides multiple scores: overall match, skills alignment, portfolio quality, and resume-portfolio alignment
  • Interview Selection - Allows recruiters to easily shortlist candidates and manage interview pipelines

How we built it

Tech Stack:

  • Frontend: Next.js 15, React 19, TypeScript, Tailwind CSS
  • Backend: Next.js API routes, Prisma ORM
  • Database: PostgreSQL (via Supabase)
  • AI/ML: Google Gemini for multi-phase agentic analysis, and for embeddings of resumes and job descriptions
  • Vector Search: Pinecone for semantic candidate matching
  • Storage: Supabase for resume file storage
  • APIs: GitHub API for portfolio analysis

Architecture: The application uses a multi-phase agentic approach for portfolio analysis, where AI agents autonomously gather context, discover relevant repositories, perform deep technical analysis, and synthesize final recommendations. Semantic search enables intelligent candidate ranking by comparing resume embeddings with job description embeddings in vector space.

Challenges we ran into

  • Designing an effective multi-phase AI workflow that balances thoroughness with speed
  • Parsing diverse resume formats and extracting structured data reliably
  • Validating resume claims against actual code in GitHub repositories
  • Building a scalable architecture that can handle bulk resume uploads
  • Creating a scoring system that fairly weighs different factors (skills, experience, portfolio quality)

Accomplishments that we're proud of

  • Built a fully functional 4-phase agentic analysis system that intelligently evaluates candidate portfolios
  • Implemented resume claim validation that cross-references stated experience with actual GitHub projects
  • Created a beautiful, responsive UI with real-time analysis progress tracking
  • Designed a semantic search system that goes beyond keyword matching
  • Developed a comprehensive scoring framework covering job match, skills, experience, and portfolio quality

What we learned

  • How to orchestrate multi-phase AI workflows for complex decision-making
  • The power of vector embeddings for semantic candidate matching
  • Effective strategies for parsing and structuring unstructured resume data
  • Building trust through transparent scoring and explainable AI recommendations
  • The importance of validating self-reported skills against tangible evidence

What's next for Lumina

  • Add LinkedIn profile analysis alongside GitHub
  • Implement team collaboration features for multi-reviewer workflows
  • Build custom scoring rubrics that adapt to different roles
  • Create ATS integrations for seamless data export
  • Add interview scheduling and email automation
  • Implement bulk candidate communication features
  • Expand AI analysis to include code quality metrics and contribution patterns
  • Add support for additional portfolio platforms (GitLab, Bitbucket, personal websites)

Built for StormHacks 2025

Built With

  • gemini
  • gemini-api
  • github
  • pinecone
  • prisma-orm-database:-postgresql-(via-supabase)-ai/ml:-google-gemini-for-multi-phase-agentic-analysis
  • react-19
  • supabase
  • tailwind-css-backend:-next.js-api-routes
  • tech-stack:-frontend:-next.js-15
  • typescript
Share this project:

Updates

Submission history