Job Type: Full-Time

ABOUT THE ROLE
Build OCR systems for printed and handwritten text in low-resource languages and complex scripts, enabling digitization of manuscripts, archives, and contemporary documents.

KEY RESPONSIBILITIES

• Develop end-to-end OCR pipelines: detection, segmentation, and recognition.
• Train and fine-tune models for handwritten text recognition (HTR).
• Adapt OCR frameworks (Tesseract, TrOCR, PaddleOCR, EasyOCR) to target languages.
• Generate synthetic data and curate annotated datasets.
• Implement preprocessing (deskewing, denoising, layout analysis) and optimize for production.

REQUIRED SKILLS & EXPERIENCE

• 2–4 years in computer vision or OCR.
• Strong Python with PyTorch or TensorFlow.
• Hands-on experience with at least one OCR framework.
• Familiarity with sequence modeling for OCR (CRNN, CTC, attention/seq2seq).
• Experience with non-Latin or complex scripts (conjuncts, diacritics, ligatures).

NICE TO HAVE

• Previous experience with Nepali language data and Devanagari script.
• Direct HTR experience.
• Document layout analysis and historical document processing.
• Synthetic data generation pipelines for OCR.
• Open-source contributions to OCR projects.

WHAT WE OFFER

• Collaborative and learning-driven work culture
• Career growth and professional development
• Competitive salary and benefits