Security research & development Accepting disclosure coordination

Tristan J. Pereira

I research and develop adversary tradecraft, evasion and detection techniques, then fingerprint whats left.

Current work

Offensive Security · AI Agents · Detection Signatures · Threat Intelligence · Infrastructure Engineering

Writing

Notes & long-form · 0

Papers

Peer-reviewed & preprint · 0
  1. 2026

    The Cheap Path to an Offensive LLM: A Local, Measured Dual-Use Capability Study

    T. J. Pereira

    Preprint · AirgapForge

Code

Selected · 0 shown · all on GitHub →

About & contact

Reports and questions welcome

I'm a security researcher working the dual-use edge of open AI models. By day I'm an AI red team specialist on a frontier-model evaluation team, where I set the technical bar for adversarial findings and judge model-generated attack tooling — functional malware, exploit code, operational uplift — against real APT-grade tradecraft.

That work builds on a decade across offense and defense: red-team operator running full-lifecycle adversary emulation and writing engagement-specific tooling in C, C#, and the Win32 API, and before that intelligence and electronic-warfare roles in the U.S. Marine Corps (GEOINT and SIGINT). The range is why my research sits where it does — what is genuinely feasible for an attacker, measured rather than assumed.

Everything here is independent work, published so defenders can reason about it and build against it. CISSP and Pentest+ certified. Reach out for research collaboration, review, or questions.