Mission
Why SecMLOps School exists.
Machine learning systems are deployed into adversarial environments with the same security assumptions as internal analytics tools. They are not the same thing. A fraud classifier behind a public API is an attack surface. A training pipeline that ingests external data is a supply chain. A model artifact stored in object storage is a program waiting to be executed.
The skills required to secure these systems sit at the intersection of MLOps, application security, and adversarial machine learning — three disciplines that rarely appear together in a single practitioner. SecMLOps School is designed to close that gap with a curriculum grounded in real threat models, real pipelines, and real incident patterns.
Every lab in the program is a milestone on a single project: an insecure fraud-detection service that students take to production-grade, control by control, with a red run and a green run as proof. No slide decks. No hypotheticals. Proven controls or it does not count.
Principles
How we build.
Practitioner-grade
Every concept is grounded in a real threat model and demonstrated with a working exploit or a working defense — not a diagram.
Proof over assertion
A control that has never been seen to fail is not a control. Every milestone requires a red run before the green run counts.
Pipeline-native
Security belongs in the CI/CD pipeline, not bolted on afterward. The program is built around GitHub Actions, GitLab CI, and Azure DevOps.
Exercises
The course includes more than 70 hands-on exercises.
Follow the build.
The program is in active development. Join the waitlist to be notified when enrollment opens and to receive early access to curriculum previews.