Security research, engineering insights, and technical deep dives

• Breaking SGLang's Trust Boundaries

  July 30, 2026 · 16 min read

  How six SGLang vulnerabilities collapsed the boundaries between inference, model loading, and distributed control.

  ai-security llm-infrastructure sglang vulnerability-research responsible-disclosure python-security cve cert-cc inference-security model-serving rce ssrf

• Khadag (खड्ग)—an autonomous offensive security agent that hunts for sandbox escapes.

  March 16, 2026 · 8 min read

  Building an autonomous offensive security agent that hunts for sandbox escapes—and solving the hallucination problem that nearly derailed everything.

  ai-security offensive-security llm autonomous-agents vulnerability-research code-sandbox khadag

• BSides Seattle 2026: Zero Trust in the Matrix

  March 5, 2026 · 5 min read

  Recap of my BSides Seattle talk on securing AI workloads on Kubernetes—covering sandbox escapes, supply chain attacks, and why your ML infrastructure probably isn't as secure as you think.

  kubernetes ai-security conference zero-trust

• 0x00: Initial Access

  March 3, 2026 · 2 min read

  Initializing the stack—what to expect from this space

  meta security engineering