Theofilos has been practicing systems engineering for 20 years, mostly in telcos, on a path that has taken him from Unix engineering to machine learning engineering. That systems background shows in how he approaches ML: as infrastructure that must be reliable, observable, and maintainable.

Today he builds tools that support companies in running their ML workloads effectively. He is passionate about Kubeflow and works across its ecosystem, including Pipelines, KFServing, Katib, and integrations with TFX, along with monitoring stacks built on Prometheus and Grafana for detecting model drift and fairness issues. His work covers the full maturity journey, from manual training runs to automated retraining in production.

Theofilos frames MLOps as culture, process, and technology rather than a pile of tools, and he shares practical guidance on notebook-to-pipeline workflows, metadata and versioning, and the trade-offs that small teams and edge deployments face.