I'm a systems engineer turned adversarial researcher. I've spent the past several years building production AI infrastructure — first as a founding engineer at Julius AI, where we scaled to over 3 million messages per week, then on developer tooling at Whop.

Today, I work independently under the Cartograph banner, running structured empirical tests against frontier models to map the attack surface of agentic AI pipelines.

My interest is in the gap between what a model is instructed to do and what it actually does when the environment is adversarial. That gap is wider than most deployment guides acknowledge.

Research Focus

• ›Prompt injection and indirect instruction attacks
• ›Multi-agent escalation and trust boundary violations
• ›Policy design and the specificity ceiling problem
• ›Capability-safety tradeoffs in frontier models
• ›Agentic pipeline defense architecture (CapabilityOS)

Links