Back to projects

Security · MLContributor

Rampart AI

Security pipeline enhancements for data exfiltration detection in production.


Highlights

  • Enhanced a production AI security pipeline by implementing Kubernetes-compatible readiness and liveness endpoints with real database and ML service verification.
  • Designed weighted pattern scoring for data exfiltration detection, differentiating bulk vs. targeted commands and ambiguous indicators to improve detection accuracy.
  • Contributed risk-based severity classification and open-sourced detection logic to the core DataExfiltrationMonitor, shipped in a production release.

Stack

KubernetesPythonML Services

Project Links