Skip to content
View KevinOBytes's full-sized avatar
7 followers · 6 following

Achievements

Achievement: Pull Sharkx3Achievement: Pair Extraordinairex3Achievement: QuickdrawAchievement: YOLO

Achievements

Achievement: Pull Sharkx3Achievement: Pair Extraordinairex3Achievement: QuickdrawAchievement: YOLO

Highlights

Block or report KevinOBytes

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
KevinOBytes/README.md

Kevin O'Connor — AI security, scientific software, and applied research

I work on AI security at EQTY Lab and build TARE through TKOResearch. OASIS is my experimental applied-science environment for hands-on work in laboratory software, hardware, and cyber-physical systems.

Earlier work included security engineering and technical direction at NSA, work on CrowdStrike's research team, and leading threat research at Adlumin and later N-able.

Writing · Tools · LinkedIn

Current work

  • EQTY Lab — AI-agent security and verifiable compute.
  • TKOResearch — applied AI-security research and scientific software.
  • TARE — a custody-aware lab-record platform connecting samples, evidence, review decisions, and exports.
  • OASIS — an experimental research environment for laboratory software, hardware, and cyber-physical systems.

Research interests

AI-agent security and tool permissions; verifiable compute; detection engineering; cyber-physical and embedded systems; scientific software and laboratory automation; local ML; and reproducible technical research.

Selected repositories

  • AgentBoundary — An open benchmark for LLM instruction-boundary behavior under prompt injection.
  • instruct — A versioned format and validator for machine-checkable software setup instructions.
  • smart-router-proxy — An OpenAI-compatible proxy that uses a local classifier to route requests by task.
  • frost — A lightweight menu-bar manager for macOS.

Selected field notes

Pinned Loading

  1. AgentBoundary AgentBoundary Public

    Open benchmark for prompt-injection robustness in LLM-based agents

    Python

  2. smart-router-proxy smart-router-proxy Public

    OpenAI-compatible proxy server with task-aware model routing — classifies requests via local DistillBERT, routes to the optimal OpenRouter model

    Python

  3. instruct instruct Public

    The .instruct file format — versioned, machine-checkable agent instruction sets for software installation and onboarding

    Python

  4. frost frost Public

    Lightweight menu bar manager for macOS Tahoe — a free, open-source alternative to Ice/Bartender

    Swift