Senior MLOps Engineer

About the role

AI71, an enterprise technology company based in Abu Dhabi that builds AI infrastructure, is hiring a Senior MLOps Engineer for its Platform Engineering team. The role builds the foundational layers that deploy, monitor and maintain models for enterprise customers, bridging model development and production. It is a platform engineering position focused on reliability, security and scalability, with an on-site arrangement in Abu Dhabi (hybrid option to be confirmed).

Key responsibilities

• Design and build the model serving layer, including real-time inference, batch processing and LLM endpoints.
• Develop CI/CD pipelines for models covering automated training, testing, versioning, staged rollouts and rollback mechanisms.
• Set up observability to track performance, model drift, data quality, cost and output quality.
• Own the evaluation infrastructure, including regression testing for non-deterministic outputs.
• Maintain infrastructure as code across all environments, meeting security and compliance standards.
• Work with data scientists and ML engineers to move from notebooks to production and define architectural standards and internal tooling.

Required profile

• Substantial hands-on experience running ML systems in production, with a focus on platform engineering.
• Strong software engineering skills in Python, with maintainable production code.
• Judgment to balance platform complexity with the company's stage, avoiding both fragile scripts and over-engineered systems.
• Collaborative approach to working with data scientists and ML engineers.

Required skills

• Kubernetes and containerization, including cluster-level architecture decisions.
• Infrastructure as code (Terraform or equivalent) and end-to-end CI/CD ownership.
• Model serving frameworks such as KServe, Seldon, BentoML, vLLM or Triton.
• ML lifecycle tooling such as MLflow, Kubeflow, Airflow or feature stores.
• Monitoring and observability with Prometheus, Grafana or similar.
• Cloud platforms (AWS, Azure, GCP or sovereign clouds).