Inspiration Job seekers spend countless hours manually tailoring resumes for each application, often missing critical ATS keywords or struggling to maintain authenticity while optimizing. We witnessed friends apply to hundreds of jobs with minimal responses simply because their resumes didn't match the specific keywords recruiters were searching for. We wanted to solve this with AI—but truthfully, never fabricating experience. What it does ResumeForge Pro is a complete multi-agent AI system that automates intelligent job matching and resume optimization: Scout Agent searches 802 indexed job postings using Elasticsearch's semantic search, scoring candidates on skills match (45%), keyword relevance (30%), experience level (15%), and competition signals (10%). Returns top 50 matches per resume. Strategy Agent intelligently filters results into APPLY/SKIP recommendations, removing overly senior roles and poor matches while prioritizing by fit score. Optimizer Agent extracts job keywords via Groq API, performs gap analysis, validates truthfulness using 40+ semantic technology clusters, and enhances bullet points naturally. Maximum 2 keywords per bullet maintains authenticity. Complete Pipeline: Automated 4-stage workflow (Scout → Strategy → Optimizer → PDF) processes resumes end-to-end in under a minute, producing ATS-friendly PDFs and beautiful HTML change logs with color-coded improvements. Results: Average +41.4% keyword match improvement, 4.8/5.0 quality rating, 100% test coverage across 65 test cases. How we built it Infrastructure (Person A):
Built PDF/DOCX resume parser handling multiple formats Designed Elasticsearch schemas for resumes and jobs with case-insensitive skill matching Implemented Scout Agent with multi-signal scoring algorithm Created Strategy Agent for intelligent job filtering Built ATS-friendly PDF generator with dynamic font scaling Developed orchestrators for complete pipeline automation
AI Optimization (Person B):
Built Optimizer Agent with Groq API integration Implemented keyword extraction and gap analysis Created truthfulness validation engine with 40+ semantic technology clusters Developed AI bullet enhancement maintaining professional tone Built comprehensive QA framework (65+ tests, 100% coverage) Created beautiful HTML changelog visualizations with color-coded keywords
Tech Stack: Python, Elasticsearch, Groq API (Llama 3.1 70B), ReportLab, PyMuPDF Challenges we ran into Preventing Fabrication: Our biggest challenge was ensuring the AI never added fake experience. We solved this by building 40+ semantic technology clusters (e.g., container_tech: [docker, kubernetes...]) that validate whether a candidate truly has related experience before adding any keyword. ATS Compatibility: Professional resume PDFs needed to match original page counts while being ATS-parseable. We implemented binary search font scaling (7-11pt) and single-column layouts to solve this. Elasticsearch Case-Sensitivity: Skills like "Python" vs "python" were treated differently. We added lowercase normalizers to the index schema, enabling truly case-insensitive matching. Real-time Performance: Processing 50+ job matches and optimizing multiple resumes needed to be fast. We optimized Elasticsearch queries and implemented efficient batch processing, achieving sub-60-second complete pipelines. Accomplishments that we're proud of
100% Truthfulness: Zero fabrication across 55+ optimized resumes—our semantic validation ensures authenticity 4.8/5.0 Quality Rating: Professional, natural-sounding optimizations that passed rigorous human evaluation Production-Ready: 100% test pass rate (65 tests), comprehensive error handling, beautiful visualizations End-to-End Automation: Single command runs complete pipeline from job search to PDF generation Real Impact: Average +41.4% keyword match improvement helps job seekers get past ATS filters
What we learned Elasticsearch is incredibly powerful: The combination of semantic search, ES|QL queries, and custom scoring enabled sophisticated job matching we couldn't achieve with simple keyword search. AI validation is critical: We learned that raw LLM output needs careful validation. Our semantic cluster approach ensures AI enhancements are grounded in reality. Testing pays off: Our comprehensive QA framework (6 major test suites) caught edge cases that would have broken production use. User experience matters: Beautiful HTML reports and clear before/after comparisons make AI changes transparent and trustworthy. What's next for ResumeForge Pro
Cover Letter Generation: Extend the pipeline to automatically generate tailored cover letters Interview Preparation: Analyze job requirements and suggest interview prep questions Application Tracking: Monitor application status and suggest follow-up timing LinkedIn Optimization: Apply same truthful enhancement to LinkedIn profiles A/B Testing: Test resume variations to identify which keywords drive best results Web Interface: Build Flask/React dashboard for non-technical users
Built With
- api
- python
Log in or sign up for Devpost to join the conversation.