Field notes

Insights from inside the fortress.

Field notes on sovereign AI for Omani institutions. Architecture, model selection, hardware sizing, and patterns of use, written by the team that builds Hosn appliances in Muscat.

Insights from inside the fortress.

Four clusters

Sovereign AI

Models

Hardware

Use Cases

Articles

On-Premise AI for Sovereign Institutions: A Complete Guide for Oman and the GCC

AI Sovereignty Under the Omani Personal Data Protection Law: A Compliance Architecture

The CLOUD Act, China's DSL, and Why GCC Sovereign Data Cannot Live in Public Cloud AI

Air-Gapped AI for Defence and Internal Security: Architecture Patterns

Sovereign AI vs Public Cloud LLM: Total Cost and Risk Comparison for Oman Government

Gemma 4 with 256K Context: A Deep Dive for Sovereign On-Premise Deployment

Qwen 3.6 for Arabic NLP: Benchmarks, Strengths, and Production Deployment

DeepSeek R1 Reasoning On-Premise: When Step-by-Step Thinking Matters

Falcon Arabic LLM: TII's Open Model and Its Place in Sovereign AI

LoRA, QLoRA, and RLHF on Customer Hardware: Fine-Tuning Without the Cloud

Choosing AI Inference Hardware: H100 vs H200 vs RTX 6000 Ada vs Mac Studio M3 Ultra

Sizing a Sovereign AI Appliance: Concurrent Users, Latency, and Throughput

Mac Studio M3 Ultra 192GB: The Surprise Sovereign-AI Edge Appliance

Strix Halo 128GB: AMD's Sovereign-AI Workstation Play

Datacenter-Grade AI Racks: Power, Cooling, UPS, and Air-Gap Network Design

AI for Defence: Arabic Document Triage and Multilingual Intelligence Workflows

AI for Sovereign Banking: Credit Memos, KYC, and AML on Private Hardware

AI for Legal Research in Oman: From Hatim to On-Premise LLMs

AI for State Audit: Anomaly Detection and Audit Copilot Patterns

AI for Healthcare in the Sultanate: Arabic Medical Scribing and Patient Notes

Aligning AI Procurement with the MTCIT Cybersecurity Framework

Reading NCSI Guidance for AI Systems in Omani Government

The Omani PDPL Cross-Border Transfer Clause and What It Means for AI

Building an AI Governance Function Inside an Omani Ministry

Patterns of AI Use Inside Police Forces: Lessons for Sovereign Deployment

How State Audit Institutions Can Use AI Without Compromising Independence

The Sovereign Wealth Fund Thesis on On-Premise AI

Sovereign-Grade AI for Tier-One GCC Banks

How Central Banks Can Use AI for Supervision Without Sharing Regulated Data

On-Premise AI for Omani Law Firms: Legal Research Inside the Office

Classified Document Workflows With On-Premise LLMs

Sovereign AI as a Pillar of Oman Vision 2040

GCC Sovereign AI Initiatives Compared: UAE, Saudi Arabia, Bahrain, Qatar, and Oman

Designing an AI Procurement RFP for the Omani Public Sector

How Public-Cloud LLM Telemetry Can Leak Classified Information

Building an AI Information Security Management System Aligned with NCSI's Second Edition

Bilingual AI for Diwan and Royal Court Correspondence

Arabic NLP for Omani Government Archive Digitization

Hyperscaler AI Services and Data Residency in Oman: What's Covered and What Isn't

AI for Parliamentary Research in the Omani Council of State and Council of Oman

Gemma 4 vs Llama 4 vs Qwen 3.6 on Arabic Evaluation

Practical Use-Cases for 256K Token Context Windows

Arabic NLP Evaluation Suites: AraBench, ALUE, and Arabic MMLU

Tokenizer Efficiency Across Arabic LLMs

Choosing Embedding Models for Bilingual Arabic-English RAG

Fine-Tuning Gemma 4 on an Omani Legal Corpus

Fine-Tuning Qwen 3.6 on Classified Documents: RLHF Safety Patterns

DeepSeek R1 for Audit Analysis Workflows

Methodology for Evaluating Multilingual Models in Sovereign Settings

Sourcing Training Data for Sovereign Fine-Tuning Without Compromising Privacy

Falcon Arabic at the Edge: Sizing for Branch Deployments

GGUF Quantization for Arabic: Q4_K_M vs Q5_K_M Quality Tradeoffs

vLLM vs TGI vs llama.cpp for Production LLM Serving

Prompt Engineering for Formal Arabic Government Tone

LLM Evaluation Frameworks for Sovereign Deployments: Ragas and DeepEval

Building an Arabic Instruction-Tuning Dataset for Sovereign Use

KV Cache Optimization for Long-Context Arabic Inference

DPO and RLHF on Private Data: An Operational Playbook

Comparing Open-Source Models on Omani Arabic Dialect

Synthetic Data Generation for Sovereign Model Training

NVLink Topology for Multi-GPU LLM Deployments

H100 vs H200 Memory Bandwidth: The Practical Impact on LLM Inference

RTX 6000 Ada 48GB for Sovereign Tower Deployments

AMD MI300X vs NVIDIA H100 for LLM Serving

Apple Silicon for LLM Inference: Benchmarks and Real-World Numbers

Edge GPU Appliances for Branch and Field Offices

Air-Gap Network Architecture for Sovereign AI Clusters

Dell vs HPE vs Supermicro AI Servers in the GCC

Power and Cooling Calculations for a 4-GPU Rack in Muscat's Climate

Choosing 2U, 4U, or Tower for Your Sovereign AI Deployment

PCIe Gen5 vs Gen4 for AI Inference: Does It Matter?

Inference Quantization vs Hardware Class: Where to Spend the Budget

LLM Concurrency Benchmarks per GPU Class