Enterprise infrastructure security at Snap. Previously Security at Microsoft. Focused on securing AI in the corporate environment, plus data loss prevention, governance and zero trust. I build open-source tooling for the same problems.
BSides Vancouver · Pattern matching stalls on messy, unstructured data. This is what LLMs add when they read for intent instead of shape.
BSides Vancouver · A hands-on look at how prompts leak. What users paste, what agents retrieve, and where the controls belong.
Compared ML models on malware detection, then tuned the best ones. The focus was hyperparameters.
Where ML helps malware classification, where it breaks, and what fixes it.
Compared the control laws that stop a 2U CubeSat tumbling.
Local AI firewall that redacts secrets and PII before prompts leave the machine. Point an agent at it with one environment variable. No CA, no system proxy. Apache-2.0. A personal project built on my own time, not affiliated with Snap.
4 HP EliteDesk mini PCs in a 10-inch rack: an OPNsense firewall, a Proxmox host and bare-metal red and blue nodes on segmented VLANs, with local SLMs for attack and triage.
IoT gateway for industrial monitoring and data collection.
Training for half-Ironmans, or on a plane chasing a swell. Most recent: Chicama, Peru.