Give your team back
30 hours a month.

Four layers. Most clients need one, and I will tell you which — an SME automating invoices and a national programme evaluating a model are not buying the same thing.

1. L1 Systems in production Agents that run real processes — documents read and posted, customers answered in Arabic and English, WPS, VAT and licence deadlines tracked — inside the tools you already use. Fixed scope, fixed price. Every action logged and human-reviewable, because an agent you cannot audit is a liability, not an asset. Peter runs this site’s front line; Mudeer is deployed at a Dubai trading company.
2. L2 Private deployment — models on hardware you control Where data cannot leave your jurisdiction, your building or your regulator’s reach, the model runs on your side of the wire instead of behind someone’s API. I run a 119-billion-parameter model in production on owned Apple Silicon — a 60-core M3 Ultra and a 128 GB M4 Max, 63 GB of weights resident locally. No inference call leaves the premises. That is the honest shape of sovereign AI at this scale: not a datacentre, but a private stack you own outright.
3. L3 The human data layer Models are only as good as the judgement they are trained against. Goldset, my AI-data venture, supplies RLHF and preference data, code and STEM annotation by screened engineers and PhDs, evals and red-teaming, Indic and Arabic multilingual data, and physical-skill video for embodied AI — to frontier teams, under a published methodology.
4. L4 Evaluation — how you know it works Rubrics built with you, gold seeded into every batch, inter-annotator agreement tracked and reported. The same discipline whether it is grading a supplier’s model or proving your own does what you claim.

What I do not do. I do not train foundation models and I do not build hardware — I buy it, run it, and know precisely what it will and will not hold. If your programme needs pre-training or silicon, I will say so on the first call rather than the fifth.

The same discipline, at AI-lab scale: YouTube — The Small Edge Goldset supplies vetted human data — RLHF, code & STEM annotation, evals and physical-skill video — to frontier AI teams. Separate venture, separate site.

AI automation, scoped and priced →

Autonomous reconnaissance against your own perimeter, then a plain-English report an owner can act on — not a 200-page scanner dump.

• Attack-surface mapping and continuous recon
• Odoo / SAP / Dynamics vulnerability assessment
• Exposed CCTV and IoT device auditing
• NESA and TDRA compliance gap reporting

How the assessment works →

From VIP banking at Noor Bank to running a family office. This is where the other three practices learned to be accountable to a number.

• Family office setup and operations
• Banking advisory and capital sourcing
• Portfolio optimisation and risk management
• RAK ICC company formation — filed direct as a licensed registered agent
• Cross-border structuring and due diligence

RAK ICC company formation — I’m the registered agent →

The full background →

Commercial electric vehicles for GCC fleets — imported, homologated and supported here, not shipped and forgotten.

• Fleet electrification strategy
• Charging infrastructure planning
• Government and municipality fleet programmes
• Commercial procurement, leasing and after-sales

Models and specifications →