Writing on AI security, network defense, and building in public.
A college-level breakdown of how DeepSeek-R1 used reinforcement learning and GRPO to incentivize reasoning, why the approach mattered, and how later work on hierarchical reasoning helps explain what may be happening inside RL-trained reasoning models.
How I added LSTM-based traffic-rate anomaly detection, decision fusion, and a dashboard to SentinelMesh.
How I trained a 1D convolutional neural network to classify IoT network traffic as benign or malicious.
How I designed a three-layer IoT intrusion detection system combining CNN classification and LSTM anomaly detection.
Why I'm building a security monitoring mesh for AI agents — and what problem it solves.
Five concepts from five years of enterprise security that show up in every serious AI incident.
My space for AI concepts, security projects, and personal perspectives on new developments in AI.