Now

Currently leading

I’m leading multiple AI and digital-transformation programmes in parallel - several mega-scale initiatives spanning citizen services, internal operations, and the platform layer that ties them together. The work is intentionally portfolio-shaped, not single-project: a flagship orchestration programme is the anchor, with a fleet of supporting initiatives running alongside it.

Across these programmes, the recurring workstreams are:

• Multi-agent orchestration architecture - moving beyond single-assistant patterns into a routing layer that coordinates specialised service agents across entities, with handoff, context, and observability handled at the platform level.
• MCP-based service onboarding - standardising how new government services expose capabilities to the AI layer so the marginal cost of bringing in another service drops sharply over time.
• RAG knowledge platform across 85+ sources - retrieval pipelines built per-source with their own freshness, access, and quality logic, composed at runtime by intent.
• Admin governance for AI lifecycle control - prompts, models, agents, and policies versioned, reviewed, and rolled out with clear gates rather than ad-hoc updates.
• Platform-vs-custom evaluation - a structured proof-of-concept programme (currently an Airia POC) to weigh platform, custom, and hybrid strategies against speed-to-market, governance fit, and long-term sustainability.
• Cross-entity stakeholder coordination - aligning multiple government entities, vendors, technical teams, and executives on shared architectural direction without any of them feeling steamrolled.
• AI Governance & Validation Framework - extending the governance discipline from prior programmes into the new portfolio so every AI capability ships with explainability, audit trails, bias monitoring, and clear KPIs.

The throughline across all of this is the same: less time on demos, more time on the operational reality that decides whether AI in government earns its place in production.

I joined Sharjah Digital Department in January 2026. Before that I spent three years at RTA Dubai (Jan 2023 – Dec 2025) leading RAG, multi-agent, and AI governance work for the 2030 Smart Mobility programme - most of the practical lessons I’m bringing into the current role come from there.

Currently writing

Two essays in active drafts:

• A field guide to multi-agent integration patterns in regulated environments - what works, what doesn’t, and where MCP fits.
• An honest look at what 40% of cancelled agentic AI projects actually have in common, drawing on Gartner’s 2025 forecast and on what I see in practice.

The first essay published on the site is “Multi-Agent AI in Government Isn’t a Chatbot Problem.” Start there if you’re new.

Currently learning

• Agno, LangGraph, and CrewAI - comparing how each handles agent orchestration, state management, and tool use under realistic governance constraints.
• Local LLM tooling at scale - what it takes to run Ollama and Llama-family models in production for retrieval-grounded answers in data-residency-sensitive environments.
• Evaluation patterns for agent systems - how to measure quality on multi-step agent flows beyond single-turn benchmarks. The literature is thin; the practice is thinner.

Currently reading

• Designing Data-Intensive Applications - Martin Kleppmann (re-reading the chapter on consistency models because it never stops being relevant).
• Build - Tony Fadell.
• PMI’s Pulse of the Profession - for the institutional view on AI adoption in project management.
• The Anthropic Model Context Protocol spec, repeatedly.

Not working on

• Cryptocurrency or blockchain anything.
• Side projects that aren’t connected to my actual delivery work.
• Open-source contributions for the next 6 months - the day job is full.

Travel

I’m based in the UAE and not planning travel. Open to visits in Dubai, Sharjah, and Abu Dhabi if you’re in the region.

Last updated May 2026. The next refresh will be when something meaningful shifts in the program or the writing focus.