Products · Applied AI
AI OCR Document Intelligence Engine
Turns dense legal bulletin PDFs into structured fields desks can confirm — then search and remind.
- Stage
- AI Infrastructure
- Users
- Via NepalIPMS desks
- AI
- Core
- Status
- Prototype → Production
What is it?
Intelligence after OCR.
Not a standalone OCR app. A staged pipeline: store pages, extract layout text, classify sections, map fields with confidence, require human review, then feed register search and deadline engines inside NepalIPMS.
Who is it for?
Operators who cannot trust silent merges.
IP practice staff who ingest Department of Industry bulletins. They need speed without abdicating judgment — every field can show provenance before it enters a matter.
Main workflows
Upload → verify → use.
UploadBulletin PDF into the firm portal
ExtractLayout OCR for Nepali / English blocks
Classify & mapIssue boundaries, application fields, confidence
ReviewHuman confirms before merge — UX of the product
Search & remindVerified data only
Screenshots & architecture
Seen in production context.
HITL
Human review before register merge
- TechnologyR2 · layout OCR · Workers · D1 · confidence scores
- UX principleConfidence and source page beside every guess
- Current statusShipped in NepalIPMS intake — not a pitch demo