ASAT Labs

ASAT Verify

Verification for customer onboarding in Nigeria.

ASAT Verify is an API that reads Nigerian identity and business documents, flags signs of tampering, and returns structured data an onboarding team can act on.

  • API-first
  • Runs in your own cloud
  • Built for Nigerian documents
Get in touch See the design
Identity documentSample
Cartoon avatar, generated illustration
NameSAMPLE HOLDER
Born · flagged01 JAN 1990
Document no.0000 000 0000
Issued01 JAN 2020
TypeID SLIP
read, checked, returned
Result
Refer for manual review
Document type
National ID slip
Fields read
5 of 5
Tamper signals
Date of birth: edited region
Cross-check
Does not match utility bill
Illustration of the intended design, using sample data and a generated cartoon portrait.

Why now

Onboarding moved to the phone. Document checks did not keep up.

Opening an account or applying for credit in Nigeria starts with identity and address documents. Those documents now arrive as photos, and the people reviewing them never hold the original.

Every account starts with a document

Banks, fintechs and lenders have to identify each customer before they open an account, so every sign-up produces documents that someone must read and judge.

Photos replaced counters

Customers send pictures from their phones. Lighting, angles and compression vary, and nobody on the team sees the physical card or certificate.

Editing an image is easy

A date, a name or an address can be changed with free tools in minutes. Manual review cannot look closely at every upload once volume grows.

Built forFintechs and digital banksMicrofinance banks and lendersOnboarding and compliance teams

How it is designed to work

Reading a document is the easy half. Noticing it was edited is the other.

Onboarding teams at fintechs, banks and lenders still key in document details by hand and judge authenticity by eye. Generic OCR misreads local document layouts and does not notice an edited image. ASAT Verify is designed around three steps.

Read

Take a photo or scan and extract its fields, with layouts tuned to the documents used in Nigerian onboarding rather than generic templates.

Check

Look for editing traces in the image, then compare fields against each other and across the documents a customer submits.

Return

Send back structured data and the reasons for any flag through an API, so a compliance team can decide what to review.

Architecture

Six stages, two of them GPU inference, all inside the customer's cloud.

This is the planned architecture for ASAT Verify. Each uploaded document passes through the same pipeline, and every flag in the result points back to the stage that raised it.

Runs inside the customer's own cloud environment
  1. CPUIntakeAccept a photo or scan, correct rotation and perspective, reject images too blurred or cropped to judge.
  2. CPUClassifyDecide which of the supported document types it is and which layout version.
  3. GPU inferenceExtractA vision-language model tuned on Nigerian layouts reads every field into structured data.
  4. GPU inferenceForensicsImage-forensics models look for spliced regions, re-compression and inconsistent fonts or spacing.
  5. CPUCross-checkRules compare fields within a document and across the documents one customer submits.
  6. APIResultReturn fields, flags and the reason for each flag as JSON, with an audit record.

Planned stack

Reading documents and detecting edits are both model inference on images, which is why the product is built around GPUs from the first prototype.

NVIDIA technology. Not in use yet. Next: run document-extraction and tamper-detection inference for the ASAT Verify prototype on rented NVIDIA GPUs, then evaluate TensorRT and Triton Inference Server for serving.

Models
An open-weight vision-language model tuned on Nigerian document layouts; image-forensics models for edit detection
Framework
PyTorch on CUDA
Compute
NVIDIA GPUs, rented by the hour for the prototype
Serving
Containerised inference service; TensorRT and Triton Inference Server to be evaluated
Deployment
Inside the customer's own cloud account, so document images stay within their boundary
Interface
REST API returning JSON, with webhooks for asynchronous results

Documents in scope

Designed for the paperwork Nigerian onboarding runs on.

The design targets six document types that customers are routinely asked for when they open an account or apply for credit.

Designed to run where the documents already are.

ASAT Verify is designed to be deployed inside a customer's own cloud environment, so identity document images stay within that customer's boundary instead of being sent to a third-party service.

Roadmap

From design to prototype to pilot.

The product design, document scope and architecture are set. The next milestone is a working prototype running on NVIDIA GPUs.

Now

Concept

Product design, document scope and architecture defined.

Next

Prototype

Run document extraction and tamper detection on rented NVIDIA GPUs with sample documents.

Then

Evaluation

Measure extraction and detection quality on a labelled test set and optimise serving.

After that

Pilot

Offer the API to a first onboarding team to evaluate inside their own cloud.

Scope: ASAT Verify is designed to examine the documents a customer presents. It does not query government identity databases and is not an identity registry.

Company

The company behind ASAT Verify.

ASAT Labs is the product brand of ADVANCED SYSTEMS & AI TECHNOLOGIES LIMITED, a company registered in Nigeria, building verification software for customer onboarding.

Legal name
ADVANCED SYSTEMS & AI TECHNOLOGIES LIMITED
Incorporated
Nigeria, July 2026
Location
Ibadan, Oyo State, Nigeria
Founder
Sikiru Uthman Ajibola
Team
Sole founder, who also builds the product
Product
ASAT Verify
Contact
sikiru@asatlabs.com