Caliber

Caliber is an AI powered Resume maker to tailor your resume as per job description.

Comment

Caliber - AI-Powered ATS Resume Builder

Inspiration

Job seekers struggle with creating resumes that pass Applicant Tracking Systems (ATS). We wanted to democratize access to professional resume optimization using free, local AI - no expensive subscriptions or data privacy concerns.

What it does

Caliber is an intelligent resume builder that:

  • Parses existing resumes from PDF/DOCX files using AI
  • Analyzes job descriptions to extract key requirements and skills
  • Generates ATS-optimized resumes tailored to specific positions with detailed scoring
  • Provides interview preparation with AI-generated questions based on your experience
  • Offers multiple professional LaTeX templates for polished, ATS-friendly output
  • Enables community learning through anonymous interview experience sharing

How we built it

Tech Stack:

  • Frontend: React + TypeScript + Tailwind CSS
  • Backend: Node.js + Express + TypeScript
  • Database: PostgreSQL with Prisma ORM
  • AI: Ollama with Gemma 2B (local, free, private)
  • Document Generation: LaTeX compilation in Docker containers
  • Queue System: Bull/Redis for async resume generation
  • Authentication: JWT-based secure auth

Architecture: Full-stack monorepo with separate frontend/backend packages, containerized services, and async job processing for scalability.

Challenges we ran into

  1. AI Model Performance: Initial model (Gemma 3:4B) was too slow (3+ min per request). Switched to Gemma 2:2B and implemented JSON format mode for 10x speed improvement.
  2. JSON Parsing Reliability: AI responses weren't consistently valid JSON. Added robust parsing with markdown removal and format enforcement.
  3. LaTeX Security: Running user-generated LaTeX posed security risks. Isolated compilation in Docker containers with timeouts.
  4. ATS Score Accuracy: Balancing keyword matching with semantic understanding required iterative prompt engineering.

Accomplishments that we're proud of

  • ✅ 100% Free & Private: No API costs, all processing runs locally
  • ✅ Production-Ready Architecture: Scalable queue system, proper error handling, comprehensive logging
  • ✅ Smart Resume Optimization: AI-powered content generation that actually improves ATS scores
  • ✅ Professional Output: Multiple LaTeX templates producing publication-quality PDFs
  • ✅ Complete Job Search Platform: Resume building + interview prep + community insights in one place

What we learned

  • Local AI (Ollama) is viable for production apps with proper model selection and optimization
  • JSON format enforcement dramatically improves LLM output reliability
  • LaTeX in Docker provides both quality and security for document generation
  • User experience matters: async processing with progress indicators prevents frustration
  • Smaller, faster models often outperform larger ones for structured tasks

What's next for Caliber

  1. Enhanced AI Features:

    • Cover letter generation
    • LinkedIn profile optimization
    • Salary negotiation coaching

  2. Advanced Analytics:

    • Track application success rates
    • A/B test different resume versions
    • Industry-specific ATS insights

  3. Collaboration Tools:

    • Resume review requests from peers
    • Mentor feedback integration
    • Team hiring workflows

  4. Mobile App:

    • On-the-go resume updates
    • Quick job application tracking
    • Interview prep flashcards

  5. Integration Ecosystem:

    • Direct job board applications
    • LinkedIn import/export
    • Calendar integration for interview scheduling
Share this project:

Updates