AI Engineer
Paperless-GPT
AI-Powered Document Analysis with Vector Search & LLM Integration
RAG
Architecture
3+
LLM Providers
<1s
Query Response
Overview
Built a document processing system that uses AI to analyze, tag, and extract data from uploads. Vector embeddings power semantic search, and LLM integration handles natural language queries against the collection.
Natural language queries like "What were last month's expenses?" return relevant documents instantly using semantic similarity rather than keyword matching.
!The Challenge
Document intelligence system needed semantic search and natural language queries against document collections.
✓The Solution
RAG pipeline with sentence transformer embeddings, ChromaDB vectors, and multi-provider LLM support (OpenAI, Ollama).
Technical Implementation
Document Processing
- PDF, image, and text document support
- OCR for scanned documents
- Automatic text extraction
- Metadata extraction (dates, amounts, entities)
Vector Search
- Sentence transformer embeddings
- ChromaDB for vector storage
- Semantic similarity search
- Hybrid search (vector + keyword)
LLM Integration
- OpenAI API (GPT-4) support
- Local models via Ollama
- Context-aware responses
- Source document citations
Tech Stack
Language
Python
Vector DB
ChromaDB
Embeddings
Sentence Transformers
LLM
OpenAI APIOllama
API
FastAPIFlask
Deployment
Docker
Skills Demonstrated
PythonChromaDBVector DatabasesOpenAI APILLM IntegrationSemantic SearchDocument ProcessingREST APIDocker