Operate & Develop ML/LLM platforms on Kubernetes + cloud (Azure; AWS/GCP ok) with Docker, Terraform, and other relevant tools
Manage object storage, GPUs, and autoscaling for training & low-latency model serving
Manage cloud environment, networking, service mesh, secrets, and policies to meet PCI-DSS and data-residency requirements
Build end-to-end CI/CD for models/agents/MCP tooling (versioning, tests, approvals)
Deliver real-time fraud/risk scoring & agent signals under strict latency SLOs.
Maintain MCP servers/clients: tool/resource definitions, versioning, quotas, isolation, access controls
Integrate agents with microservices, event streams, and rule engines; provide SLAs, tracing, and on-call runbooks
Measure operational metrics of ML/LLM (latency, throughput, cost, tokens, tool success, safety events)
Enforce governance: RBAC/ABAC, row-level security, encryption, PII/secrets management, audit trails.
Partner with DS on packaging (wheels/conda/containers), feature contracts, and reproducible experiments.
lead incident response and post-mortems.
Drive FinOps: right-sizing, GPU utilization, batching/caching, budget alerts.

4+ years in DevOps/MLOps/Platform roles building and operating production ML systems (batch and real-time)
Strong hands-on with Kubernetes, Docker, Terraform/IaC, and CI/CD
Practical experience with Spark/Databricks and scalable data processing
Proficiency in Python & Bash
Ability to operate DS code and optimize runtime performance.
Experience with model registries (MLflow or similar), experiment tracking, and artifact management.
Production model serving using FastAPI/Ray Serve/Triton/TorchServe, including autoscaling and rollout strategies
Monitoring and tracing with Prometheus/Grafana/OpenTelemetry; alerting tied to SLOs/SLAs
Solid understanding of PCI-DSS/GDPR considerations for data and ML systems
Experience with the Azure cloud environment is a big plus
Operating LLM/agent workloads in production (prompt/config versioning, tool execution reliability, fallback/retry policies)
Building/maintaining RAG stacks (indexing pipelines, vector DBs, retrieval evaluation, hybrid search)
Implementing guardrails (policy checks, content filters, allow/deny lists) and human-in-the-loop workflows
Experience with feature stores
Qwak Feature Store, Feast
A/B testing for models and agents, offline/online evaluation frameworks
Payments/fraud/risk domain experience; integrating ML outputs with rule engines and operational systems
Advantage
Familiarity with Databricks Unity Catalog, dbt, or similar tooling

Machine Learning Operations Engineer

Key skills