Field notes from our engineering practice
Deep, opinionated writing on the work we do every day: production AI, retrieval, DevOps, FinOps, and digital transformation programmes.
Building bilingual English and Arabic products: what right-to-left really costs
Bilingual delivery in the Gulf is not a translation task at the end of the build. Right-to-left layout, mixed-direction content, Arabic typography, and locale-aware data touch the design system, the components, and the test suite. Plan for it from the first sprint.
AI-augmented operations: where an SRE agent earns trust
An agent that can read your telemetry, runbooks, and pull requests can cut time-to-detect and take toil off the on-call rota. The line between helpful and dangerous is the line between proposing and executing, and it has to be drawn in policy, not in prompts.
Offline-first field apps for inspectors and frontline teams
Inspection, enforcement, and maintenance teams work in basements, on sites, and at borders where connectivity is unreliable. Apps that assume a network fail exactly when the work matters. Offline-first is an architecture, and it changes the data model.
Market intelligence agents: from document firehose to explainable signals
Research and trading desks drown in filings, news, and alternative data. Multi-agent systems can turn that into ranked, cited signals, but only if explainability is designed in from the first line. Unexplained scores do not survive compliance.
Billing anomaly and dispute agents: finding the problem before the customer does
High-volume billing produces errors at a rate that hand-tuned thresholds either drown in or miss. Anomaly agents that learn what normal looks like per account, paired with dispute agents that triage consistently, turn billing from a complaint channel into a control.
Sub-second AI on live sports data: what the latency budget actually buys
Broadcast overlays and live briefs have to land while the moment is still on screen. That budget rules out most hosted model round-trips and forces a specific architecture: streaming ingest, compact models at the edge, and frontier models reserved for what can wait.
Designing operator consoles for high-stakes workflows
Case-management, inspection, and surveillance consoles are used for hours a day by people whose mistakes have consequences. Designing them like consumer apps produces polished screens and slow, error-prone work. Here is what we do instead.
Writing architecture decisions that outlive the programme
Most architecture documentation is written for the steering committee and read by nobody after sign-off. Decision records written for the engineer who inherits the system in three years are a different artefact, and they change how programmes behave.
Sovereign cloud is a design constraint, not a deployment target
Data residency and classification rules shape everything from identity to observability. Treating sovereign cloud as a late-stage deployment choice is how programmes end up rebuilding half their platform. Design for it from the first diagram.
Building production RAG on local LLMs
What it actually takes to run retrieval-augmented generation on open-source models inside your own estate: model selection, hardware sizing, retrieval design, evaluation, and the operational practices that keep it honest.
AI cost optimisation for cloud workloads: an agent-led approach
Cloud cost growth keeps outpacing engineering capacity to manage it. Agentic FinOps closes the gap by combining recommendation agents, guarded execution, and the chat-ops loops engineers actually use.
Designing AI compliance agents for regulated industries
Compliance AI fails when it sounds confident and lacks evidence. The architectures that work in regulated industries are built around citations, audit trails, and human authority, not model-of-the-month leaderboards.
Want to talk about any of this?
The articles are field notes from real engagements. The conversations are where they get useful.