MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202611058314 A) filed by Mr. Pravin Kumar Bansal; and Dr. Brijendra Gupta on May 07, 2026, for On-Prime Ai Processing, Conversion And Preservation Of Archived Documents Agnostic Of Format And Language To A Structured Database.
Inventors include Mr. Pravin Kumar Bansal; and Dr. Brijendra Gupta.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: The present invention provides a method, system, and non-transitory computer-readable medium for the on-prime (on-prime) AI-driven ingestion, processing, conversion, and structured preservation of archived documents agnostic of format, physical condition, script, or language, into a validated structured database. The invention comprises eight sequential pipeline stages: format-agnostic ingestion, image pre-processing and enhancement, on-prime OCR extraction, document classification, schema resolution, Vision-Language Model extraction, validation and confidence scoring, and structured output to an on-prime database. The classification-first design resolves the correct output schema before field extraction commences, enabling schema-guided extraction rather than generic mapping. All processing is performed exclusively on on-prime GPU hardware with no transmission of document data to any external network or cloud service, rendering the system suitable for air-gap-compatible, security-sensitive deployment environments. The system achieves greater than 95% document classification accuracy and greater than 85% handwritten field extraction accuracy, with per-record confidence scoring and a human review queue for records below the confidence threshold. The invention supports documents in any language or script, domain-adaptive fine-tuning via LoRA adapters, and dynamic registration of new document types. It is industrially applicable across defence, healthcare, government, judiciary, finance, and any large-scale document archive conversion programme. FIG 01.
Disclaimer: Curated by HT Syndication.
