Patents & Publications

Public patent applications, technical articles, and media connected to production machine learning work.

Patents

Source Code Programming Language Prediction

US20240086187A1

Machine learning system for classifying source-code programming language from source-code artifacts.

Google Patents

Behavior-Based Asset Classifications

US20250202921A1

Behavior-based classification methods for cybersecurity assets using observed activity and contextual signals.

Google Patents

Session-Based Anomaly Detection on Cloud Logs

US20250233875A1 / EP4586117A1

Operational prediction on user-based contextual sessions, including anomaly detection on cloud log activity.

US filing EP filing

Multi-Modal Breach Detection with Time-Aware Graph ML

US20260095470A1

Cybersecurity breach prediction using multiple data modalities and graph-based machine learning methods.

Google Patents

False-Positive Prevention with Multi-Modal Data and Graph ML

US20260089177A1

Prediction of false-positive cybersecurity detections using multi-modal analysis and graph ML.

Google Patents

Interpretable ML-Based Alert Grouping for NextGen SIEM

US20260111548A1

Interpretable cybersecurity detection grouping for analyst triage and alert investigation workflows.

Google Patents

Technical Articles & Media

AI-Powered Risk Scoring with Falcon Next-Gen SIEM CrowdStrike · Feature video

Data science lead for the ML-powered risk scoring system that helps analysts prioritize high-priority threats in Falcon Next-Gen SIEM.

Watch video

How CrowdStrike Achieves Lightning-Fast Machine Learning Model Training with TensorFlow and Rust CrowdStrike Engineering Blog · June 2022

Technical article on reducing ML model training time from 227 hours to 6 hours through Rust-based feature extraction, GPU infrastructure, and TensorFlow input-pipeline optimization.

Read article

New CrowdStrike Research Challenges Container Predictability Assumptions CrowdStrike Engineering Blog · October 2024

Research article analyzing billions of container events across popular applications and testing assumptions about workload predictability.

Read article