ResumeGPT
overview.
A modern web application that lets you create, edit, and optimize professional resumes using natural language prompts and AI-powered suggestions. Just describe your experience or role, and ResumeGPT helps you turn it into a polished resume. Choose from multiple templates, live preview your changes, and export your resume as a PDF in seconds. Featuring advanced ATS analysis, semantic skill matching, and real-time AI voice interviews powered by LiveKit and Sarvam AI.
technical implementation.
Recruiter Matchmaker & Client-Side Embeddings
Built a semantic candidate search engine using client-side ONNX inference (Xenova/bge-small-en-v1.5) via Web Workers. Embeddings are securely stored in Postgres with pgvector and queried using HNSW indexing for rapid cosine similarity matching.
AI Integration with Groq & Llama 3.3
Integrated Groq API running Llama 3.3 for lightning-fast natural language processing, converting conversational input into structured resume data, generating tailored cover letters, and providing deep RAG insights for recruiter matchmaking.
Advanced ATS Analysis with RAG & NLP
Implemented an ATS compatibility engine using Retrieval-Augmented Generation (RAG) and NLP. Features semantic skill extraction, bidirectional skill matching, TF-IDF keyword analysis, and fuzzy string matching (Jaro-Winkler) for robust evaluation.
Real-time AI Voice Interviews
Integrated LiveKit and Sarvam AI to provide interactive, low-latency AI-driven voice interview simulations. Uses @livekit/components-react for real-time communication and Groq for the voice-to-text-to-voice reasoning loop.
Robust Authentication & Data Management
Implemented secure Google OAuth authentication using NextAuth.js with PostgreSQL persistence via Prisma ORM. Features per-user isolated candidate pools, session management, and resume data versioning.
Serverless PDF Generation Pipeline
Built an enterprise-grade PDF generation system using Puppeteer Core and @sparticuz/chromium optimized for Vercel serverless environments, creating high-quality, ATS-compatible resume exports.
key features.
- AI-powered resume content suggestions using Groq (Llama 3.3)
- AI Cover Letter Generator for tailored, instant cover letters
- Recruiter Matchmaker with semantic candidate search and Groq RAG insights
- Advanced ATS compatibility analysis with Local RAG and NLP
- Real-time ATS scoring and optimization suggestions
- Semantic skill extraction and bidirectional matching
- Smart keyword extraction using TF-IDF analysis
- Fuzzy skill matching with Jaro-Winkler algorithm (85% threshold)
- Real-time AI voice interviews with LiveKit and Sarvam AI integration
- 10+ professional ATS-optimized resume templates
- Live editing and real-time preview functionality
- High-quality PDF export with @sparticuz/chromium
- Google OAuth authentication with NextAuth.js
- Chat session management and persistence
- Responsive dark/light theme UI with Framer Motion animations
screenshots.
AI-powered resume building with natural language input
Add your own Gemini API key to use your quota instead of the shared one.
Chat naturally and see live preview with download option
Choose from Modern, Minimal, Classic and 10+ more templates
ATS analyser specific page of user custom resumes
ATS analysing of generated resume using ResumeGPT
challenges & solutions.
Challenge: Gemini AI responses needed consistent structured output for resume data
Solution: Engineered comprehensive prompt engineering with strict JSON schema validation, custom parsing logic, and error handling to ensure reliable resume content generation with proper data types and formatting
Challenge: PDF generation in serverless environments faced memory and timeout constraints
Solution: Migrated from standard Puppeteer to @sparticuz/chromium with optimized configurations, implemented efficient rendering pipeline, and added proper error handling for reliable PDF generation in Vercel environment
Challenge: ATS compatibility while maintaining visual appeal across multiple templates
Solution: Researched ATS parsing requirements and implemented template designs that balance visual aesthetics with machine readability, using proper semantic HTML structure and optimized formatting
Challenge: Ensuring accurate ATS scoring and skill matching for diverse job descriptions and resume formats
Solution: Developed a robust NLP pipeline with RAG, TF-IDF, and fuzzy matching (Jaro-Winkler) to handle variations in job descriptions and resume content, ensuring reliable semantic skill extraction, bidirectional mapping, and real-time compatibility analysis across multiple industries.