My client is one of the most recognisable tech platforms in the Middle East, operating across 70+ cities in 10 countries, serving tens of millions of customers, and now making a serious push into AI. They're looking for engineers who can turn that ambition into production-ready systems.

The Role

You'll sit within the Data Science team, working on GenAI applications that directly shape the product experience at scale. This isn't research for research's sake, you'll be building and deploying real systems used by millions of users across the region.

What You'll Be Doing

• Building and deploying RAG systems and LLM-powered applications in production
• Designing scalable GenAI pipelines using LangChain and LCEL
• Running randomised controlled experiments and translating results into product decisions
• Implementing observability and tracing for LLM-powered systems in production
• Collaborating with product, data science, and engineering teams to ship AI-driven features
• Driving exploratory analysis to identify new product levers and opportunities
• Challenging existing approaches and keeping the team aligned with industry best practices

What They're Looking For

• 4–6 years in machine learning and/or software engineering
• Advanced degree in a quantitative field (CS, Engineering, Mathematics, Statistics, Physics)
• Strong working knowledge of transformer architectures, attention mechanisms, and LLMs
• Hands-on experience with advanced prompting — CoT, few-shot, in-context learning
• Proficiency in LangChain/LCEL for building complex LLM workflows
• PyTorch or TensorFlow
• FastAPI or similar for building scalable APIs
• Proficient in Python, SQL, Spark/Hive
• Solid software engineering practices — Git, CI/CD, automated testing
• Distributed training knowledge (DeepSpeed, Megatron-LM) is a bonus

What You Get

• Work on AI problems at genuine scale — tens of millions of users across a complex, multi-market platform
• Be part of a team leading the AI ecosystem in the MENA region
• A platform that's actively investing in its next chapter, not just maintaining what exists