Machine Learning

Purpose of the Position: The AI Platform Engineer- MLOps/LLMOps will be responsible for designing, implementing, and operationalizing enterprise-grade AI, Machine Learning, and Generative AI solutions. The role will focus on deploying and monitoring AI applications, establishing governance frameworks, and enabling secure, reliable, and cost-effective AI operations across cloud environments. The individual will provide technical leadership and drive best practices for AI engineering, MLOps, and LLMOps initiatives.

Key Result Areas and Activities:

• Reliability and Availability of AI services and solutions
• Design scalable AI/ML and Generative AI deployment architectures.
• Define enterprise standards for model deployment, monitoring, and lifecycle management.
• Design and operationalize RAG, agentic AI, prompt management, vector databases, and LLM integration frameworks.

• MLOps & LLMOps Implementation
• Establish CI/CD pipelines for machine learning and LLM-based applications.
• Automate model training, deployment, versioning, and rollback processes.
• Optimize LLM performance, scalability, and cost efficiency including Inference cost management

• Model Governance & Reliability
• Implement frameworks for model monitoring, evaluation, drift detection, observability, and compliance.
• Ensure responsible AI and governance standards are followed.

• Technical Leadership & Stakeholder Collaboration
• Mentor AI engineers, junior platform engineers and data scientists.
• Collaborate with business and technology teams to translate AI use cases into production-ready solutions.

• Essential Skills:
• LLM serving and inference optimization and running an inference gateway that routes across providers and open-weight models with failover
• LLM observability and tracing: end-to-end request tracing through retrieval, prompt, tool calls and generation using LangSmith or Azure AI Foundry
• Evaluation as a platform capability: building the harness product teams use to regression-test AI behaviour, run LLM-as-judge and offline evals, and gate releases on measured quality rather than judgement calls
• Prompt, model and config lifecycle management: versioning prompts and system messages like code, safe rollout and rollback, A/B and shadow testing of model or prompt changes in production
• Token and inference cost engineering: spend attribution by team, feature and tenant; caching and model-routing strategies; GPU utilisation and right-sizing
• Retrieval infrastructure operations: running and tuning vector or hybrid search at scale, embedding pipelines, index refresh strategies, and retrieval quality monitoring
• Kubernetes and GPU infrastructure in production: Docker, autoscaling, node pools, GPU scheduling and sharing, plus infrastructure as code (Terraform or Bicep) and GitOps (Argo CD or Flux)
• Strong Python and the software engineering discipline to build self-service platform tooling and internal SDKs that engineers actually adopt
• Classical MLOps foundations: model registry and versioning (MLflow, Azure ML, or SageMaker), automated retraining and redeployment pipelines, drift and data quality monitoring, and pipeline orchestration with strong SQL