Applying research to solve real-world problems — improving handwritten Tamil recognition, and the evaluation work that tells me whether it actually improved.
My research bridges software engineering and language technology, building practical solutions that improve accessibility, document digitisation, and regional language computing.
I approach research as an industry engineer, with a strong emphasis on reproducibility, practical implementation, and measurable improvement. The work combines traditional Optical Character Recognition with modern AI to better recognise handwritten and printed Tamil documents.
The goal is technology that preserves knowledge, improves accessibility, and supports the digital transformation of regional language content.
Improving recognition accuracy for handwritten and printed Tamil documents through better datasets, model training, preprocessing, and evaluation methodology.
Technologies that support the digital ecosystem of the Tamil language — from encoding through to preservation.
Integrating large language models into OCR pipelines to correct errors, restore missing characters, recover formatting, and improve readability. AI serves as a post-processing layer that improves extracted text — it does not replace the OCR engine, and it must not invent text that was never on the page.
Improving access to printed and handwritten Tamil content for visually impaired users through OCR-powered assistive applications.
Improving handwritten Tamil optical character recognition using Tesseract OCR and AI-assisted post-processing.
Large-scale dataset preparation and ground truth generation for handwritten Tamil, including the tooling to make that work repeatable.
Training and fine-tuning OCR models, then optimising Character Error Rate against a held-out evaluation set.
Character Error Rate (CER) and Word Error Rate (WER) analysis, to measure whether a change actually helped.
Post-processing pipelines that use language models to clean up OCR output without inventing text that was never on the page.
This research feeds directly into Vizhi Tamil, the accessibility app that puts Tamil OCR in users' hands.
An engineering-driven methodology, start to finish.
Step 01
Step 02
Step 03
Step 04
Step 05
Step 06
Published work, datasets, and experimental results.
A technical textbook written to give students a structured introduction to computing fundamentals. I served as first author; the experience strengthened my interest in technical education and knowledge sharing.
An open benchmark for Tamil OCR covering handwritten and printed text — roughly 769,396 file pairs (TIFF images, ground truth, and box files) totalling 3.6 GB. Source text is drawn from Wikipedia, Wikisource, Maattru, and Theekkathir, rendered across 27 Unicode fonts: 9 handwritten and 18 printed, from Google Fonts and Tamil Virtual University. Released CC0, with data collection by the Kaniyam Foundation.
Machine translation matters most for low-resource languages, and Tamil is hard: complex morphology, syntax, and cultural nuance. This paper evaluates Claude, ChatGPT, and Gemini on English-to-Tamil translation across culturally sensitive, political, technical, and idiomatic content, scored with BLEU, BERT-based metrics, METEOR, and TER. Gemini ranked highest on accuracy and precision for complex syntax and idioms, ChatGPT led on semantic alignment, and Claude was strongest on document-level fluency.
SyedKhaleel Jageer, Priyaradhikadevi T., Madhan K., Prasanna S. Panamerican Mathematical Journal, Vol. 35, No. 2s (2025).
My published research covers OCR evaluation methodologies, dataset generation, Tesseract optimisation, AI-enhanced OCR, machine translation, Tamil language computing, and accessibility. It is indexed on ResearchGate and Zenodo.
2023
2024
2025
2026
Research should solve real problems. I believe meaningful research is practical, reproducible, measurable, open, accessible, and useful to people — measured not only by publications, but by whether the solutions built from it are any good.
I welcome opportunities to work with researchers, academic institutions, open-source communities, and industry partners across OCR, artificial intelligence, Tamil language technologies, accessibility, Android development, and open source software.
Research becomes meaningful when it transforms ideas into technologies that improve people's lives.