Daris Chen

Overview

PatentIQ AI is an intelligent patent search and drafting assistant built to simplify patent research for students, inventors, and early-stage founders. The platform combines semantic search capabilities with LLM-based analysis to help users identify related patents, understand the competitive landscape, and receive AI-assisted drafting suggestions. By removing barriers to professional-level patent research, PatentIQ AI democratizes access to patent information for those who need it most.

(GitHub Link)

Quick Summary

AI-powered semantic patent search using vector embeddings and PostgreSQL pgvector for meaning-based retrieval rather than keyword matching.
Idea submission and analysis system that accepts text or document uploads and queries USPTO patent databases to identify related, pending, or approved patents.
Hybrid ranking system combining semantic similarity (70%) and keyword matching (30%) to deliver most relevant patent results.
LLM-based recommendation engine that analyzes user submissions and provides structured suggestions for patent development and drafting.
Keyword expansion and CPC classification suggestions to improve search recall and provide industry context.
JWT authentication with Auth0 for secure user management and personalized research tracking.
Full-stack application with FastAPI backend, Next.js frontend, and Supabase database integration.

Key Features

Semantic Patent Search
- Vector embeddings convert patent text and queries into numerical representations
- PostgreSQL with pgvector enables efficient similarity-based retrieval
- Search finds patents by meaning, not just exact keywords
- Interactive terminal-based search interface with real-time results
Hybrid Ranking System
- Combines semantic similarity scores with keyword matching
- Configurable weighting (default: 70% semantic, 30% keyword)
- Returns top-ranked patents based on relevance to user query
- Provides similarity scores for transparency and decision-making
Idea Submission & Analysis
- Accept text descriptions or document uploads of invention ideas
- USPTO API integration for live patent database queries
- Identify existing patents that may conflict or relate to the idea
- Detect pending applications and approved patents in similar domains
Recommendation Engine
- LLM analyzes user submissions against retrieved patents
- Generates structured recommendations for next steps
- Identifies patent gaps and opportunities for filing
- Suggests related technology areas and CPC classifications
AI-Assisted Drafting
- Provides suggestions for patent specification content
- Helps draft claims based on similar existing patents
- Offers language suggestions aligned with patent office conventions
- Supports iterative refinement of draft documents
CPC Classification Suggestions
- Automatically suggests Cooperative Patent Classification codes
- Maps user ideas to appropriate patent categories
- Helps understand competitive landscape by classification
- Guides searchers to most relevant patent sections
User Authentication & Profile Management
- JWT tokens with Auth0 integration
- Secure user accounts and research history tracking
- Personalized recommendations based on search history
- Privacy-first design with encrypted data storage