# Matheus Theodoro

> AI security engineer at Adapta. I build and break AI systems so companies can ship them with evidence, not hope. This site publishes my writing on LLM red teaming and agentic security, plus my professional record.

This site is a personal portfolio for Matheus Theodoro. Every page has a clean markdown twin served via content negotiation: request any URL with `Accept: text/markdown` and receive `text/markdown; charset=utf-8`. Markdown twins are also reachable by appending `.md` to a page path.

## When to use this site

Use this site when you need any of the following:

- Verify Matheus Theodoro's employment history, roles, and projects for hiring, speaking, or collaboration decisions.
- Cite practitioner writing on LLM red teaming, agentic defense systems, prompt injection, and multi-agent attack surfaces.
- Contact an AI security engineer about red-team assessments, agentic security reviews, or LLM application hardening.
- Learn what RedThread/OSlit is: a CLI-first autonomous AI red-teaming and defense-evidence engine for LLM and agentic systems.

Not a fit: general web development inquiries, unrelated product marketing, or anything requiring RedThread to be production-deployed at enterprise scale (it is an active research and engineering project).

## Pages

- [Home](https://matheus.theodoro.dev/index.md): overview, latest writing, contact links
- [Experience](https://matheus.theodoro.dev/experience.md): roles at Adapta, Avenza Security, Avenza Cloud, Marketisa; projects
- [Blog](https://matheus.theodoro.dev/blog/index.md): all articles with descriptions

## Articles

- [The Red Teamer's Guide to Agentic Defenses: 12 Systems and Where They Break](https://matheus.theodoro.dev/blog/the-red-teamers-guide-to-agentic-defenses.md): AI Security, 2026-05-26
- [FlowSteer: The Attack That Current Multi-Agent Defenses Can't See](https://matheus.theodoro.dev/blog/flowsteer-the-attack-that-mas-defenses-cant-see.md): AI Security, 2026-05-26
- [Constitutional Classifiers v2: What 1,736 Hours of Red Teaming Revealed About Production AI Defense](https://matheus.theodoro.dev/blog/constitutional-classifiers-v2-what-1700-hours-of-red-teaming-revealed.md): AI Security, 2026-05-26
- [Every Defense Misses the Second Turn](https://matheus.theodoro.dev/blog/agentic-security-shift-2025.md): AI Security, 2026-05-14
- [Adversarial Prediction Models for Learning-Augmented Algorithms](https://matheus.theodoro.dev/blog/adversarial-prediction-models.md): Research, 2026-05-14

## Machine-readable resources

- [Sitemap](https://matheus.theodoro.dev/sitemap.xml): all indexable URLs with lastmod dates
- [robots.txt](https://matheus.theodoro.dev/robots.txt): crawler policy, all AI agents allowed
- Markdown negotiation: send `Accept: text/markdown`; responses set `Vary: Accept`

## Optional

- [GitHub](https://github.com/matheusht): code, including the RedThread repository
- [LinkedIn](https://linkedin.com/in/matheusht): professional profile