AI Security, Red Teaming, and Secure Deployment

Jonathan R. Reed

I find out how AI systems fail before their users do, then build the controls, deployment patterns, and developer tools that make them safer.

I study Sociology at UT Dallas because model failures rarely stop at the model. Users, incentives, and operating systems decide what happens next, especially when an AI system can retrieve private data or act through tools.

Open to internship and entry-level AI security, secure deployment, and software roles. Dallas or remote.

View work View resume Contact
GitHub

Selected work

All projects

Pick the right resume

Same history, different emphasis: separate versions for safety research, security engineering, and software roles. Grab the closest fit.

Resume hub

Waves

tools

A native macOS per-app audio mixer for volume, EQ, output routing, and adaptive focus without a virtual audio driver.

Waves 1.2 is public, open source, signed and notarized, and distributed as a universal DMG or Homebrew cask.

Visit project

RAGFuzz

security

A local-first workspace for testing retrieval systems against injection, poisoning, leakage, and multi-turn failures.

Saved runs produce redacted HTML, Markdown, and JSON evidence that can be inspected, compared, and replayed locally.

Read case study

Strata Studio

creative

A browser studio that turns real terrain and OpenStreetMap data into generative poster art.

A live studio can regenerate real places into reproducible, exportable compositions while proxying terrain through a bounded Worker.

Visit project

Featured AI security writing

Browse all posts