Orfeas K.

Senior ML Platform Engineer. I build and operate the production ML platforms that turn research models into reliable enterprise services, and most recently, the agentic AI layer on top of them.

For the past 4 years I've been the senior ML Platform Engineer behind Universal Music Group's core AI/ML estate. I designed and run the multi-tenant inference platforms that serve UMG's audio AI models at catalog scale.

Lately I've been deploying UMG's foundation audio models to production via a custom inference API, exposing them as tools through an MCP server, and wiring up enterprise agentic workflows (production-grade LLMOps).

• End-to-end MLOps platform ownership — async/batch inference on AWS (SageMaker, Lambda, EventBridge, S3), orchestrated by Airflow + Kubernetes/EKS, Terraform-managed infrastructure, GitHub Actions/Tekton CI/CD.

• LLMOps & Agentic AI - production MCP server design, tool-routing, agent orchestration, and OAuth OBO for user-scoped permissions across the agent surface. The patterns for wiring foundation models into enterprise systems safely.

• Secure multi-tenant ML systems - Vault-backed secret management, tenant isolation, and audit trails for ML over rights-restricted or regulated data.

• Platform engineering depth - observability, model versioning, reproducibility, zero-downtime deploys, IaC.

Differentiator: PhD in Electrical Engineering and a deep DSP/audio-ML background, so I'm equally comfortable writing high-performance inference runtimes as I am wiring up an MCP server, Airflow DAG or a Terraform module — a combination most MLOps generalists can't offer.

Looking for: senior MLOps / ML platform contracts, 100% remote across EU/UK, 3–6 month engagements with extension potential. Available evenings or weekends to gauge potential with prospective clients before engaging on a 5-day contract.