AI Architect – Azure OpenAI Platform & Safety

Theomnihire; Abu Dhabi; On-site; AED 350,000 - 520,000; Full time

Job summary

Theomnihire in the United Arab Emirates is seeking a senior AI/LLM product leader to design and deploy responsible AI features on Azure OpenAI in UAE regions. The role focuses on building a safeguarded AI gateway, enabling explainable suggestions, and ensuring high standards for data redaction and audit trails.

You will craft guardrails, evaluation methods, and a content pipeline to convert large slide decks into CMS-ready English and Arabic content, with human verification at each unit.

Qualifications

• 10+ years in software with leadership in AI/LLM product areas.
• Hands-on retrieval, prompt and tool design, evaluation, guardrails and cost control.
• Experience with Azure OpenAI and Azure platform in regulated settings.
• Knowledge of data residency and privacy obligations in UAE.

Responsibilities

• Design AI features on Azure OpenAI in UAE regions, including an AI CV builder with explanations.
• Set up and run AI gateway with data redaction and thorough request logging.
• Build safety for minors: boundaries, refusals, red-teaming and evaluation harness.
• Develop content pipeline turning hundreds of slides into CMS-ready English/Arabic content, with verification.
• Define delivery team usage of AI review tools and policy guardrails.
• Advise on governed citizen development for client staff in later phases.

Skills

LLM product design

Prompt design

Tool design

Evaluation guardrails

Azure OpenAI

Azure platform

Data privacy

Cost control

AI risk communication

Agentic coding tools

Tools

Claude Code

GitHub Copilot

Job description

Theomnihire in the United Arab Emirates is seeking a senior AI/LLM product leader to design and deploy responsible AI features on Azure OpenAI in UAE regions. The role focuses on building a safeguarded AI gateway, enabling explainable suggestions, and ensuring high standards for data redaction and audit trails.

You will craft guardrails, evaluation methods, and a content pipeline to convert large slide decks into CMS-ready English and Arabic content, with human verification at each unit.