Backed by Prosus, VentureSouq, Dash Ventures and Sanabil 500. $3.6M raised to power the future of lending.

Five years of credit infrastructure, on day one

Orbii Labs is a credit modelling environment, a workspace where your credit team understands its own data, engineers credit features, trains and optimises models, and promotes a champion into production.

Book a demo
Orbii Labs SME working capital · study v4
Credit teamBuild a working capital model on the last 18 months.
OrbiiGraphed 6 tables, 214 columns from POS, core banking and bureau. I'd start from three lenses: cashflow volatility, supplier concentration, repayment rhythm.
Credit teamAdd our own lens for Ramadan seasonality.
OrbiiDone. Trained 3 candidates. v4 leads and holds drift under threshold across the last 6 cohorts. Validation report ready to export.
The four workspaces all driven from the chat
01Data Graphs6 tables, 214 columns, profiled and joined
02Credit Studio4 lenses, 3 ours and 1 yours, 37 features
03Holistic BandingA to D, with the drivers behind each
04ML Studio3 candidates trained, pipeline frozen
ABCDbands Championv4 Gini0.62 Backtested18,400 loans
Nothing runs live until your team promotes it. Promote to decisioning
Trusted by Risk Leaders at

Slowed by Reviews, Stuck with Risky Guesses

Manual underwriting process slows down decisions
Limited ability to confidently approve thin-file SMEs
Credit teams stretched across spreadsheets, PDFs, and rules engines
Difficult to scale loan book without compromising risk

Three steps.

One champion model.

01

Reads your tables

02

Engineers the features

03

Trains and validates

01 · Data graphs

It reads your tables before it proposes anything

Column profiling, written descriptions of each table and its granularity, schema relationships, and functional clusters generated from your own data rather than preconfigured. This is the foundation everything else is built on.

Deterministic profiling across every column

Clusters generated from your data, not a template

Table quality and granularity surfaced up front

Data graphs6 tables · 214 columns
Profiled from your own schema
pos_transactions1 row per transaction · 4.1M rows98% filled
core_loan_master1 row per loan · 18,400 rows100% filled
bureau_enquiries1 row per enquiry · sparse61% filled
erp_ar_ageing1 row per invoice · 902K rows94% filled
Functional clusters found
Cashflow volatilitySupplier concentrationRepayment rhythmIdentityCollateral
Generated from your data. Nothing here was preconfigured.

02 · Credit studio

If your data can't support a feature, it says so

Credit lenses are the angles you judge a borrower from: revenue stability, operational activity, industry benchmark. Every requested feature is tested against your actual schema, and one that isn't computable is returned as not feasible rather than approximated.

A lens library spanning invoice financing, working capital and more

Features generated per lens, with the code exposed

Add and refine lenses of your own

Credit studioLens: working capital
Features tested against your schema
Cashflow volatility, 90dcomputable
Supplier concentrationcomputable
Repayment rhythmcomputable
Ramadan seasonality, yourscomputable
Inventory turnsnot feasible
No stock table exists in your schema, so inventory turns is returned as not feasible rather than approximated from something adjacent.
37 features generated across 4 lenses. The code is exposed for every one.

03 · Machine learning studio

The AI designs the experiment. It does not do the fitting.

Cross-validation and hyperparameter sweeps run deterministically. The agent reads the results and designs the next experiment, the work a data scientist would otherwise do by hand. Import a model you already trust and it becomes the baseline the challenger is built on.

Deterministic training and validation

Experiment comparison and champion selection

Bring your own model as the seed

ML studio3 candidates trained
Gini on holdout18,400 loans
v2 baseline, yours0.54
v3 challenger0.58
v4 champion0.62
v4 leads and holds drift under threshold across 6 cohorts.Promote
Nothing runs live until your team promotes it.

Make Credit a Growth Engine,
Not a Bottleneck

92% ↓

in credit evaluation time

+68%

in approvals

58% ↓

in default rate

How Foodics Lender boosted loan approvals by 30% using Orbii’s AI engine.

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut et massa mi.

More

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut et massa mi.

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut et massa mi.

More

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut et massa mi.

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut et massa mi.

More

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut et massa mi.

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut et massa mi.

More
Designed for developers

One record, one API.

A single API and one webhook stream across intake, decisioning, servicing and portfolio. Every field traces to its source, every decision is logged.

LN-00001
POST /v1/applications
{
  "borrower": "Falcon Foods LLC",
  "documents": [ 4 files ],
  "policy": "sme_v3"
}
// 201 Created
{
  "loan_id": "LN-00001",
  "status": "parsing",
  "events": "webhook"
}
documents.parsed 312 fields from 4 documents09:41:02
application.assembled screened against sme_v309:41:04
decision.issued Band B, PD 4.1%09:41:06
loan.disbursed on approval—
repayment.posted on each instalment—
portfolio.updated continuously—
policy.applied sme_v3, by rules engine09:41:04
model.scored champion v4, Gini 0.6209:41:06
reason.recorded DSCR 1.42, zero days past due09:41:06
field.traced every value back to its source page09:41:06
reviewer.assigned credit officer, queue sme09:41:07
Every event carries the same loan id. Nothing is re-keyed between surfaces.

Built for credit teams
that move fast

Train and validate models on your own book

Backtest a champion before anything goes live

Nothing runs live until your team promotes it

Live across UAE, KSA, Oman and Jordan

Trusted by risk leaders

Our modelling team works alongside yours

Book a demo