البنوك والتقنية المالية· BANKING & FINTECH · RISK & DATA SCIENCE

Basirahبصيرة

Risk, AML, and credit models that actually read Arabic transaction memos.

A risk and data-science platform for banks and finance companies — credit-decisioning, AML detection, and customer segmentation models trained on real Saudi transaction data, including the Arabic free-text memos most vendor models silently discard.

الأرقام تكذب أحيانًا، والنمط لا يكذب

“Numbers sometimes lie. The pattern never does.”

What Basirah does

Arabic-native transaction understanding

NLP models trained on Saudi transaction memos and merchant descriptors — not translated English models retrofitted onto Arabic text.

Explainable credit decisioning

Every score ships with the weighted factors behind it, ready for a human underwriter or a SAMA audit to review in one screen.

AML typology detection

Structuring, rapid circular transfers, and cross-border red flags are surfaced and scored before they become a filed STR.

Behavioral segmentation

Continuously re-scored customer segments replace static annual reviews, so risk tiers move as behavior actually changes.

Drift and bias monitoring

Every production model is watched for data drift and demographic bias, with automatic recalibration alerts before accuracy degrades.

Sovereign model hosting

Training data and model weights never leave Kingdom infrastructure — PDPL-aligned and NDMO-classified from day one.

See Basirah in action

An early UX walkthrough of the product, designed bilingually from day one.

1Portfolio risk dashboard screen — neuriya Basirah

Portfolio risk dashboard

2AML alert queue screen — neuriya Basirah

AML alert queue

3Credit decision detail screen — neuriya Basirah

Credit decision detail

How it works

01

Connect

Plug into core banking, card, and payment-rail feeds without duplicating your existing data warehouse.

02

Train

Models are trained and validated on your own transaction history, including Arabic memo and merchant text.

03

Decide

Credit, AML, and segmentation scores flow into underwriter and compliance workflows with full explainability.

04

Monitor

Drift, bias, and performance are tracked continuously, with recalibration proposed before accuracy slips.

SAMA CSFPDPLNDMO ClassificationISO 27001

Stop retrofitting English models onto Arabic risk

We're selecting design-partner banks and finance companies for the first pilot cohort. Book a briefing to see the models on your own portfolio data.

Request a Pilot