A Python FastAPI service providing two APIs for a logistics platform: a multi-prompt chatbot powered by OpenAI GPT for HS Code lookup, port codes, and CEISA query assistance — and an OCR API for extracting data from shipping documents using OpenAI Vision.
- Built the entire FastAPI service from scratch — application structure, router separation (chat, upload_ocr), environment config, and deployment setup with uvicorn.
- Implemented the multi-prompt chatbot API supporting multiple system-role prompts per request, enabling domain-specific behavior tuning without code changes.
- Built the OCR API for uploading images and PDFs and extracting structured text and data via OpenAI Vision for logistics document processing.
- Managed GPT model configuration with gpt-4o-mini as the default inference model, with response format tuning for structured logistics data (HS Code, port code, CEISA responses).
- Integrated image preprocessing using pillow and pytesseract for pre-Vision OCR fallback, and refactored system prompts for consistent logistics domain accuracy.