AI Governance Engineer

Company Overview

Talabat is the premier on-demand platform for food and non-food delivery across the MENA region, active in eight countries and handling hundreds of millions of orders each year. As part of Delivery Hero, a global leader in online food delivery and quick commerce, Talabat emphasizes an engineering-first approach.

Role Summary

The AI Governance Engineer will take full ownership of establishing standards, performing testing, and independently validating the safety, quality, and compliance of our expanding AI agent fleet. This role does not involve building or operating AI agents but focuses on defining what it means for agents to be safe, correct, and compliant, developing scalable tests, and enforcing accountability when issues arise. The engineer will implement risk-tiering and reusable testing frameworks to manage growth efficiently.

Key Responsibilities

• Design and maintain adversarial payload libraries and red-teaming frameworks utilized by engineering to self-assess AI agents.
• Create reusable red-teaming libraries containing payload templates and attack patterns addressing OWASP top vulnerabilities for large language model applications and Talabat’s specific agent use cases.
• Facilitate autonomous verification of data leakage prevention and tool guardrails by engineering teams through standardized testing suites.
• Lead periodic red-team campaigns and develop automated regression scripts to empower teams in managing their safety posture.
• Develop hallucination testing policies, bias assessment templates, and drift monitoring frameworks accessible as services.
• Integrate evaluation methodologies into continuous integration and deployment (CI/CD) pipelines to embed quality assurance in the development process.
• Implement automated triggers to enable spot-checking of production conversations to uphold high-quality standards.
• Translate regulatory compliance requirements into actionable technical checklists and mapping tools for independent execution by engineering teams.
• Establish frameworks for auditability and brand safety sampling to operationalize compliance in real-time.
• Bridge the gap between legal and engineering through technical verification tools that facilitate governance as an enabler instead of a bottleneck.
• Oversee automated monitoring and reporting of governance metrics to maintain a comprehensive overview of all AI products.
• Conduct periodic maturity assessments aligned with global frameworks such as NIST AI RMF and the EU AI Act, and develop roadmaps to address compliance gaps.
• Collaborate with teams to resolve issues and meet target governance metrics when performance falls below standards.

Candidate Profile

The ideal candidate possesses extensive experience in building and testing systems where safety and compliance are paramount. Experience in red-teaming or adversarial testing specifically with large language model applications, AI engineering, QA automation, compliance engineering, or security testing is essential. A strong understanding of data access control models—including IAM, row-level security, and policy-based access control—is required. Familiarity with retrieval-augmented generation (RAG) pipelines and agent tool-calling mechanisms is beneficial. Proficiency in coding, preferably Python, to develop test harnesses, automate scans, and analyze logs is necessary.

Desirable Attributes

Successful candidates will fluently communicate across security, engineering, data, and legal domains, adeptly translating technical concepts between stakeholders. They should be comfortable handling technical challenges and articulating trade-offs between governance rigor and development speed. Prior experience implementing governance standards and compliance workflows in regulated sectors such as fintech, healthtech, financial services, or government is advantageous. Knowledge of prompt injection defenses, OWASP top vulnerabilities for LLMs, governance frameworks, and the ability to make decisions independently under ambiguity are highly valued.

Career Progression

• Initial Phase: Conduct a comprehensive audit of all agents, define governance taxonomy, perform initial red-team campaigns, build cross-functional relationships, and establish a workflow for findings.
• Operationalization: Roll out reusable red-teaming libraries, automated regression tests, data leakage standards, hallucination evaluation frameworks, and brand safety reviews.
• Scaling & Stratification: Develop risk-tiering systems, extend automated red-teaming capabilities, build self-service governance libraries, and deploy drift monitoring dashboards.
• Embedded & Strategic: Achieve fully automated governance with no silent drift incidents, ensure governance is viewed as an enabler rather than a barrier, and influence safety-first design principles among teams.