What we put into production.

01The challenge

Citizens moved between fourteen disconnected portals to complete a single request, re-entering the same data at each step. Peak-season load regularly stalled the oldest systems, and there was no single record of a request as it crossed departments.

02What we built

A unified services platform with one identity, one inbox and one audit trail across every department. We built an Arabic-first, fully RTL interface, a workflow engine that routes each request to the right desk, and a public status tracker that shows applicants exactly where they stand.

03How it’s engineered

An event-driven core fronted by a versioned API gateway, with each legacy system wrapped behind an adapter so departments could migrate one at a time with zero downtime. Data is encrypted in transit and at rest, access is least-privilege and fully audited, and the whole stack runs with UAE data-residency options. Load is absorbed by horizontal autoscaling proven against simulated peak-season surges.

04Outcomes

• −71%

  average request time

• 99.98%

  platform uptime

• 1.6M

  requests handled / year

01The challenge

Dispatch ran on phone calls and end-of-day spreadsheets. Operators could not see where vehicles were in real time, exceptions surfaced hours late, and reporting to clients was assembled by hand each morning.

02What we built

A real-time control tower that ingests telematics from the fleet, maps every shipment to its vehicle, and raises exceptions — delays, route deviations, idle time — the moment they happen. Dispatchers reassign in a drag, and clients get a self-service portal with live tracking.

03How it’s engineered

A streaming pipeline normalizes device telemetry into a single event store; a geospatial service computes ETAs and geofence breaches continuously. The UI is a websocket-driven dashboard kept smooth under thousands of moving markers via tiling and clustering. Built mobile-first for in-cab tablets, with offline-tolerant sync over patchy coverage.

04Outcomes

• −43%

  idle fleet time

• 12s

  exception-to-alert latency

• 3,400

  vehicles tracked live

01The challenge

Onboarding was a manual chain of document uploads, email handoffs and back-office checks. Drop-off was severe, approvals took days, and every regulatory change meant re-coding the flow.

02What we built

A configurable onboarding engine where compliance rules are data, not code. New applicants move through identity capture, risk scoring and approval in one guided flow, while the compliance team adjusts thresholds and steps from an admin console without a release.

03How it’s engineered

A rules engine separates policy from implementation, so requirements change by configuration. Verification providers sit behind a swappable adapter layer; every decision is written to an immutable, tamper-evident audit log. Secrets are managed centrally, sensitive fields are encrypted at the field level, and the flow was hardened with in-house penetration testing before launch.

04Outcomes

• −92%

  time to approval

• +38%

  completed signups

• 0

  critical findings at launch

01The challenge

Buyers across dozens of stores forecast by gut feel against last year’s numbers. Fast movers sold out while slow stock tied up cash, and there was no shared view of what was actually selling where.

02What we built

A forecasting assistant that learns each store’s rhythm — seasonality, local events, promotions — and recommends replenishment per item, per branch. Buyers ask it questions in plain Arabic or English and get an explained answer, not just a number.

03How it’s engineered

A nightly pipeline consolidates point-of-sale and inventory feeds into a clean feature store; forecasting models run per store-item and are continuously back-tested against actuals. A retrieval layer grounds the natural-language assistant in the group’s own data so answers are explainable and traceable. Everything runs on managed cloud with cost guardrails and clear data lineage.

04Outcomes

• −34%

  stock-outs

• −21%

  excess inventory

• +11%

  gross margin