Sovereign AI-Assisted Software Engineering in Government

Context

Engineering teams want AI coding help. A government organisation cannot send its source code to outside services to get it. This program gives engineers AI coding help while keeping the code inside the organisation.

What I led

Sovereign AI-assisted development program

I lead a sovereign AI-assisted development program:

• Self-hosted open-weight coding LLMs on government-controlled cloud GPU, with IDE coding agents, so source code never leaves the organisation.
• AI usage guidelines and governance aligned with UAE Information Assurance data classification.
• AI-assisted development training, delivered to engineering teams.
• Secrets and credentials hardening across 3 codebases (Android, iOS, backend).

The program targets 100% of repositories scanned and remediated, zero plaintext secrets in new commits, 80%+ of engineers onboarded within 3 months, and 99%+ platform availability.

Agentic “ticket-to-pull-request” assistant

I authored the business case for an agentic engineering assistant that takes a ticket through to a pull request. It is still in planning. The design is an AI coding agent with RAG over the codebase, a human approval gate, and automated pull requests.

• A 7-week, 4-phase plan with go/no-go gates.
• Targets automating 40–60% of routine tickets.
• Estimated savings of ~100 engineering hours a month and ~USD 120K a year.

Local MCP server for enterprise email

I designed and built a local MCP server that connects an AI assistant to enterprise email with read and draft-only access. It processes everything on the device, in line with UAE data-residency requirements.

Why it matters

Sovereign AI is not about avoiding AI. It is about where the AI runs and who controls the data. Treating data classification, secrets hygiene and human approval gates as part of the design is what lets a government engineering organisation use AI at all.