These are the projects and systems that best represent where my work is now: research-informed, operationally grounded, and built to be used under pressure.
Research + Benchmark
Co-authored paper built from the largest public AI agent red-teaming competition to date, turning 1.8 million prompt-injection attacks into concrete evidence of persistent deployment risk.
Public Repo
Elixir/OTP SIEM and NDR platform with telemetry ingestion, correlation, historical storage, and terminal-first investigation workflows.
Public Repo
Purple-team platform extended with MCP, Hermes-facing boundaries, and bounded Jido-powered investigation orchestration.
Public Repo
Ongoing home-lab build with a Raspberry Pi 5 OpenWrt edge router, dual GbE HAT, managed switching, VLAN design, and separate infrastructure, AI, pentest, IoT, and guest segments.
Public Repo
Technical dataset and resolver workflow for identifying real media types behind mislabeled DOJ archive files under WAF and rate-limit constraints.
Currently Private
Phoenix and LiveView desktop-OS style environment with a built-in SIEM demo, dual-surface application design, and reusable domain boundaries.
Currently Private
Public-facing Phoenix LiveView site for nonprofit workflows with Slack-delivered forms, embedded validation, and abuse-aware request handling.
Currently Private
Modular Elixir command-and-control core exploring concurrency, pluggable transports, hot-loaded plugins, and red-team infrastructure design.
Currently Private
Phaser-based side-scroller with an AI-powered intercom and jailbreak puzzle that turns prompt attack concepts into an interactive security experience.