Multi-stage deep learning pipeline that parses scanned handwritten calendars into structured data — segmenting regions, classifying day/month cells, and recognizing handwritten annotations.

OCR/HTR web service for digitizing Moscow church records (18-19th century). Extracts structured data (names, dates, addresses) from historical documents with NER and WER quality metrics. Hackathon project for Moscow Main Archive.

AI-powered document scanner that automatically detects, corrects perspective, and enhances scanned documents from photos using OpenCV

"Interactive OCR system with real-time correction using Tesseract and confidence-based filtering."