Associate AI Quality Engineer

Who Are We❓ We Are Foodics! a leading restaurant management ecosystem and payment tech provider. Founded in 2014 with headquarter in Riyadh and offices across 5 countries, including UAE, Egypt, Jordan and Kuwait. We are currently serving customers and partners in over 35 different countries worldwide. Our innovative products have successfully processed over 6 billion (yes, billion with a B) orders so far! making Foodics one of the most rapidly evolving SaaS companies to ever emerge from the MENA region. Also Foodics has achieved three rounds of funding, with the latest raising $170 million in the largest SaaS funding round in MENA, boosting its innovation capabilities to better serve business owners. The Job in a Nutshell💡 You'll build the internal AI systems our engineers work inside every day: agents that generate and maintain tests, pipelines that triage failures before a human sees them, tooling that speeds up code review and debugging, and the evaluation infrastructure that makes our own AI features testable at all. This is a builder's role. You'll write production code, own systems in CI, and be measured on whether engineers actually use what you ship. What Will You Do❓ Test automation across the stack - Backend: API and contract testing, service-level and integration coverage, data setup that doesn't rot, and test design that survives a schema change. - Frontend: web E2E and component-level coverage with Playwright, visual and RTL regression, and suites fast enough to gate a merge rather than a nightly. - Mobile: native and cross-platform coverage with Appium or Maestro, device-farm strategy, offline and sync behaviour, and the payment-peripheral paths that only break on real hardware. - The connective tissue: shared fixtures, environment and test-data management, parallelisation, and CI pipelines where a red build means something. - performance and load testing Experience. The AI layer on top of it - Test generation from specs, code, and production traffic — with the maintenance story solved, not just the first draft. - Failure triage that classifies a red build before a human opens it: real bug, flake, environment, or test rot. - Self-healing locators and suite health tooling — flake detection, quarantine, coverage-gap analysis. - Evaluation infrastructure for AI features across our products: datasets, scoring, and regression detection when a prompt or model changes. - Evaluation for our market specifically — Arabic and English behaviour, RTL interfaces, and region-specific POS, tax, and payment rules. Correctness here is rarely a string match. Agentic AI and orchestration - Agentic AI that does real work in our pipelines: reads a diff, runs the relevant suite, reproduces a failure, proposes a fix, opens the PR. - Agents that own a quality workflow end to end — exploratory testing against a running build, coverage-gap hunting, release-risk assessment — and know when to escalate to a human. - Orchestration that holds up under load — multi-step planning, tool use, retries, state and memory across steps, sandboxed execution, multi-agent handoffs, and clean boundaries between agentic and deterministic steps. - Integration with the stack we already have (CI, Jira, observability, MCP-style tool interfaces) rather than a parallel system beside it. - The judgement to know when a plain pipeline beats an agent, and to say so. The technical ground You should be current on how this work is actually done today, and able to argue about it rather than recite it: - Test automation: framework design and layering, the test pyramid and where it stops being useful, flake economics, parallel execution, mobile and cross-browser realities, CI/CD gating, Framework: Playwright, Appium, Maestro - Agentic AI: orchestration and tool use, multi-step planning, memory and state, sandboxed execution, multi-agent patterns, MCP and similar tool-integration standards, and the cost of each. - Context engineering: retrieval strategy, chunking, reranking, caching, and managing long-context behaviour — including where it degrades. - Evaluation: offline and online evals, LLM-as-judge and its failure modes, human-in-the-loop review, statistical significance on small samples, regression gates in CI. - Reliability: structured output, guardrails, fallback and retry design, and handling non-determinism in systems that must not flap. - Operations: tracing and observability for LLM systems, prompt and version management, latency and cost budgeting, model routing, and when fine-tuning or distillation beats a better prompt. What Are We Looking For❓ - An engineer who ships production software, with recent hands-on work on LLM-backed systems that real users depend on. - Strong Python; comfortable in at least one of .NET, Java, or TypeScript. Tested, maintained code — not notebooks. - Real automation depth across more than one surface. You've owned a suite that gates releases on backend and on a UI — web or mobile — and you can explain how yo

An engineer who ships production software, with recent hands-on work on LLM-backed systems that real users depend on. Strong Python; comfortable in at least one of .NET, Java, or TypeScript. Tested, maintained code — not notebooks. Real automation depth across more than one surface. You've owned a suite that gates releases on backend and on a UI — web or mobile — and you can explain how you approach these challenges.