Consultants who ship
Slides don’t run in production. Our teams write code, stand up infrastructure, and stay until the system carries real load — the roadmap and the repository come from the same people.
AI-native consulting · Riyadh · كندي
AI that ships.Impact that stays.
Kindi takes Gulf enterprises and government from AI ambition to production systems — agentic engineering, LLM platforms, and Arabic-first AI — and leaves behind capability, not dependency.
Strategy → production
Engagements end in running systems, not slide decks.
Arabic-first
RTL-native, dialect-aware AI for 400M+ speakers.
Sovereign by design
Data residency, PDPL, and government-grade delivery.
Capability transfer
Your team runs it after we leave. That’s the point.
01 — What we do
Five practices that turn AI budgets into working systems.
- 01
AI Strategy & Roadmap
A costed, sequenced build order your board can fund and your engineers can start.
- 02
Agentic Systems
Agents that move real work through real systems — with evals, guardrails, and audit trails.
- 03
LLM Platforms
Governed, cost-controlled LLM infrastructure on your cloud or on-prem — pilot to platform.
- 04
Arabic-first AI
AI that performs in Arabic as well as English — dialects, RTL, and evals included.
- 05
Capability Building
Academies, playbooks, and an operating model — so the tenth use case doesn’t need us.
02 — Why Kindi
Most AI initiatives stall between the deck and the deployment. Ours don’t.
أثرٌ يبقى
Arabic is a first language
Most AI treats Arabic as a translation afterthought. We engineer for it from day one — dialect handling, RTL interfaces, Arabic evals — because your customers and citizens live in it.
Built for this market
Data residency, PDPL compliance, national-cloud deployment, government procurement — we design around Gulf constraints from the first architecture diagram, not as a retrofit.
Built for scrutiny
Trust & SecuritySovereign-ready
National cloud, sovereign regions, or on-prem — residency decided at the first diagram.
PDPL-aligned delivery
Saudi data law treated as a build requirement: consent boundaries, lineage, subject rights.
Your IP, entirely
Code, models, and evals land in your repositories. We keep nothing to rent back.
Audit-ready by construction
Every model decision logged, traceable, and explainable — before the regulator asks.
03 — Where we work
Regulated sectors are our home ground.
04 — How we work
Four phases. One direction: production.
01
Diagnose
Two to three weeks inside your organization. We leave with a costed roadmap of the use cases worth building — and the ones to kill.
02
Prove
A production-grade pilot on real data in six to eight weeks, measured against baselines you sign off before we write code.
03
Industrialize
The pilot becomes a platform: security review, evals, observability, cost controls, and integration into the systems you already run.
04
Transfer
Your engineers ship the next use case with our playbooks. Dependency on us going down is a KPI we report.
05 — Insights
Field notes from production.
Playbook
From pilot to platform: the 70% nobody demos
The distance between a working pilot and a governed production system is where most enterprise AI budgets die. A field guide to the unglamorous majority of the work.
June 2026 · 7 min read
Perspective
The agentic enterprise: what actually changes
Beyond the hype cycle, agentic systems change three specific things about how work moves through an organization. The rest is continuity.
May 2026 · 5 min read
Technical
Arabic is not a translation target
The performance gap between English and Arabic is invisible in the vendor demo and everywhere in production. What Arabic-first engineering actually requires.
April 2026 · 6 min read
06 — Before you ask
The questions every buying committee asks.
Straight answers, in writing, before the first call — because procurement shouldn’t have to dig for them.
Usually with a Readiness Sprint: two to three weeks, fixed fee, ending in a costed roadmap and a kill list. It’s the cheapest way to find out what’s real — and it de-risks everything that follows.
You do. Code, models, evaluation suites, documentation — everything lands in your repositories under your IP from day one. We keep nothing back to rent to you later.
Inside your perimeter. We design for data residency from the first architecture diagram — national cloud, sovereign regions, or on-prem — and PDPL obligations are treated as build requirements, not legal footnotes.
Senior-only teams who write the code they recommend, milestone pricing instead of hourly burn, Arabic engineering as a first-class discipline, and engagements designed to end — with our own redundancy reported as a KPI.
No. We’re vendor-neutral: models, clouds, and tooling are chosen against your requirements and benchmarked on your workloads — and we show the eval results that justified every choice.
Fixed fee for sprints, milestone-based for builds, retainers for embedded partnership. Never open-ended time and materials — an engagement without a defined end isn’t a service, it’s a dependency.
Bring us a problem worth solving.
A 45-minute call with our team — no pitch deck, no juniors. We’ll tell you what AI can and cannot do for your organization, and what we’d build first.