Skip to content
Smart Lead Tech
Home
Services
Our work
Saudi Arabia
About us
Blog
Contact us
🇸🇦العربيةContact now
Smart Lead Tech

Smart software solutions

Follow us

Company

  • About us
  • Our work
  • Our services
  • Saudi Arabia
  • Restaurant POS
  • Contact us
  • Book a call

Quick links

  • Blogs
  • Pricing
  • Refund policy
  • Privacy policy
  • Terms and conditions
  • Data deletion

Contact us

  • Egypt — Head Office

    Office 6, Bldg 23, Makkah Al-Mukarramah St., Omraneya, Giza 3540203, Egypt
  • Saudi Arabia — Branch

    Bldg 2646, Secondary 123, Street 7697, Ishbiliyah Dist., Jubail 35416, Kingdom of Saudi Arabia
  • Email address

    help@smartleadtech.com
  • Phone number 🇪🇬

    +20 102 201 4671
  • Phone number 🇸🇦

    +966 53 044 7112

© 2026 Smart Lead Tech. All rights reserved.

  • Smart Lead Tech Information Technology LLC·Commercial Register ⁦61239⁩·Tax ID ⁦774-391-316⁩

  • Smart Lead Tech Software Establishment·CR ⁦7054902585⁩·VAT

343 pages and 99.42%: what building SLT OCR taught us about Arabic documents

When we started SLT OCR we thought the problem was “reading Arabic letters”. After a year of invoices, purchase orders, ID cards and KYC packs, it turned out reading was the easy part. This is an honest post-mortem of wh…

09/05/2026 — 11 min read

An Arabic invoice next to a field-extraction screen showing confidence scores

When we started SLT OCR we thought the problem was “reading Arabic letters”. After a year of invoices, purchase orders, ID cards and KYC packs, it turned out reading was the easy part. This is an honest post-mortem of what worked and what failed, with numbers.

The number we are proud of — and what is behind it

On a production sample of 40 documents (343 pages): 99.42% page accuracy and 97.5% whole-document accuracy. The gap between the two is the whole story: one error on one page drops the entire document out of the “correct” column. So the goal was never higher reading accuracy, but knowing when we are wrong.

Lesson one: a verification model per field, not one model for everything

A tax number obeys a mathematical checksum. A date must be a real date. A grand total must equal the sum of the lines. When every field got its own validator, the system became able to say “I am not sure” — the most valuable sentence in document processing.

A value that cannot be verified is not delivered as fact; it goes to a human review screen. That is not a feature, that is product ethics.

Lesson two: Arabic is a layout problem, not a glyph problem

  • Mixed columns: Arabic description, English code, and a price in Eastern or Western digits on the same line.
  • Stamps and signatures over text — far more common on Gulf paperwork than you would think.
  • “Scanned” PDFs shot on a phone, skewed, with the owner’s hand shadow.

The fix was not a stronger OCR model but a layout-normalisation stage before reading, and per-document-type lexicons (a Saudi invoice ≠ an Egyptian invoice).

Lesson three: privacy is an architectural decision

Documents are never written to disk; processing happens in memory, and sources live in a 15-minute window before disappearing. The on-premises edition is a Docker container with zero network access at runtime — models and lexicons baked into the image. Banks and government bodies did not ask about accuracy first; they asked about this.

Lesson four: the export is the product

Nobody wants “pretty JSON”. They want an Excel workbook with stable keys that drops straight into their accounting system, and one HTTP API for automation. When we connected the output to the Tawqe3k e-signature platform, the path became complete: scanned document → verified data → signed contract, with no manual re-entry.

What we would do differently

  1. Start with a real test set from the client before writing any code.
  2. Measure whole-document accuracy, not character accuracy.
  3. Build the human review screen in week one, not the last week.

Document AI is not an accuracy race; it is a trust race. And trust is built by a system that knows its limits and says them out loud.

Share

Projects related to this article

  • AI SystemsPythonNext.jsDocker

    SLT OCR — Arabic-first Document Data Extraction

    An Arabic-first computer-vision engine that turns invoices, purchase orders, receipts, ID cards and KYC packs into data you can trust — a verification model per field, human review when needed, processed in memory only.

    Show more
    SLT OCR — product landing on desktop and phone
  • Real EstateLaravelVue.js

    Mashrouk — Real-Estate Co-Investment Platform

    A platform that brings individuals together to co-own and co-invest in real estate — securely and transparently from listing to closing.

    Show more
    Mashrouk — organised real-estate co-investment
  • Real EstateLaravel

    Real Estates Matcher — AI-Powered Property Matching

    A smart real-estate platform that matches demand and supply using AI and data analytics.

    Show more
    Real Estates Matcher — smart demand-supply matching

Related Blogs

  • A digital identity verification flow: ID card, live selfie and Nafath

    Know your customer in 90 seconds: how eKYC became the baseline in the Gulf and Egypt — Nafath, liveness and document AI

    Three years ago “opening an account” meant a branch, a form and a signature. Today, any financial, real-estate or even crowd-investing product that cannot verify a customer’s identity within two minutes on a phone loses …

    09/13/2026 — 10 min read

  • An integration map between an app and Saudi government platforms

    An engineer’s guide to Saudi government platform integrations: Nafath, Absher, Wathq, Ejar, Sakani — what repeats in every project

    Every serious product in Saudi Arabia reaches the same moment: “we need to integrate with Nafath”. Then Wathq for commercial registrations, Ejar for contracts, then Sakani, Absher or Muqeem depending on the sector. We ha…

    09/08/2026 — 10 min read

  • A retinal scan next to a traffic-camera frame with detection boxes

    Computer vision outside the lab: two stories from the road and the eye clinic

    Models look dazzling in the slide deck. The truth shows up at noon on a dusty highway, or in a clinic that sees eighty patients a day. These are two stories about the gap between “98% on the test set” and “actually works…

    05/19/2026 — 9 min read