Projects Resume

GitHub LinkedIn

Projects - Computer vision

American Ship OCR / Computer Vision

Segmentation demo from the package-label detection pipeline.

Image segmentation customer label detector.

2024-2025Shipped internallyPythonYOLOv8Google Tesseract OCRLangChainRAG

Problem

Package inbounding depended on slow manual reading of shipping labels and barcodes under company-specific imaging conditions.

Built

Trained a YOLOv8 image segmentation model to locate shipping labels and barcodes.
Fine-tuned Google Tesseract OCR for the company dataset, reaching 95% text extraction accuracy.
Built a LangChain RAG assistant for common customer questions and explored database-backed retrieval.

Impact

Reduced package processing time to 2 seconds per package.
Improved OCR quality over the default Tesseract model.

Edgar Yepez

Email GitHub LinkedIn Projects