A student building a form-digitisation tool for a local school found that OCR that worked well on printed textbooks fell apart on handwritten admission forms. Handwriting recognition is a genuinely different problem, not a harder version of the same one.
Why it's harder
- Handwriting varies wildly between writers; printed text is drawn from a small set of consistent fonts.
- Forms in India often mix Devanagari and Roman scripts on the same line — a single-script model misses half the content.
- Scanned or photographed forms add skew, shadows and creases that clean printed-text datasets never include.
A practical approach
Use a vision-capable LLM for handwritten extraction rather than a classic OCR engine trained on printed fonts — modern multimodal models handle messier handwriting meaningfully better. Always keep a human-review step for any field that feeds something consequential, like marks or ID numbers.
See designing an OCR pipeline for PDFs for the printed-text baseline this builds on.
GenAI engineer and AI product builder with 2+ years shipping production-grade AI systems: RAG pipelines, fine-tuned LLMs, hybrid retrieval and multi-modal apps across vision, speech and OCR, architected end to end from ingestion to deployment. Leads engineering for SCULT INDIA's 14-member team, founded the 500+ member TechVerse Enclave community and has mentored 200+ students. Three hackathon first prizes: Changethon 2025 (IIT Roorkee), Product Genesis at Vividhotsava 2025 (CSJMU Kanpur) and BYTEBATTLE (MeetKats).
Open to GenAI roles, hackathon judging, mentorship sessions and guest talks at colleges. Email pranjulrathour41@gmail.com.
