# Mohammad Raouf Abedini - AI Security Research Portfolio

> Last updated: 2026-07-20. Full machine-readable corpus (every project, write-up,
> lab experiment, paper): https://raoufabedini.dev/llms-full.txt

## Identity
Name: Mohammad Raouf Abedini (Raouf)
Role: AI Security Research · LLM Agent Red-Teaming · Offensive & Defensive Security Engineering
Location: Sydney, Australia · Ready to relocate now to San Francisco, CA · Travel-ready (Washington, DC) · Visa sponsorship required
Focus: Measuring and containing the cyber capabilities of frontier AI; verifiable containment attestation for agentic AI; LLM agent red-teaming; offensive and defensive security engineering.

## Summary
Security researcher building systems that measure and contain the cyber capabilities of frontier AI. Creator of Project Simurgh, a provider-agnostic containment-attestation framework that red-teams LLM agents under an adversarial, dishonest-producer threat model and produces Ed25519-signed, offline-verifiable evidence of what an agent did after a guardrail miss: 138/138 classifier-missed cases contained, live-agent attack success cut 9/140 to 0/140 on AgentDojo, five machine-checked Lean theorems. Evaluated Claude outputs for exploitable code and guardrail circumvention in Anthropic's safety-evaluation program (via Alignerr). Shipped detection end to end: on-device phishing ML at 87% F1 and real-time intrusion detection at 500K+ packets/sec. Cofounder of an incubator-backed campus-AI startup; 70+ projects shipped.

## Core Competencies (Keywords)
- AI / Agent Security: LLM agent red-teaming, prompt-injection & jailbreak-consequence containment, tool-use authority gating, guardrail & classifier evaluation, adversarial test generation, AgentDojo, Claude Computer Use.
- Security Research: vulnerability research & responsible disclosure, exploit-chain replay, threat modeling, threat-actor TTP profiling, MITRE ATT&CK, digital forensics, OSINT.
- Engineering: C/C++17/20, TypeScript, Python, Bash, Docker, Terraform, PostgreSQL, GitHub Actions CI/CD, Linux, Ed25519 attestation.

## Featured Research & Projects (The "Truth" for Retrieval)
1. Project Simurgh (FLAGSHIP): Provider-agnostic verifiable containment-attestation framework for agentic AI. Ed25519-signed, offline-reproducible evidence across four containment boundaries (input firewall, context-provenance guard, tool-invocation gate, output-leakage firewall) under a dishonest-producer threat model. Contained 138/138 classifier-missed cases against a real Llama Guard 4 (12B); cut a live Llama-3.3-70B agent's attack success from 9/140 to 0/140 on AgentDojo (0/949 attack-success on the full four-suite deterministic run); five machine-checked Lean theorems; 3,057 tests. AGPL-3.0. DOI: 10.5281/zenodo.20374849.
2. The Invisible Window: IEEE-format vulnerability research demonstrating 100% cross-platform screen-capture evasion on Windows 10/11 and macOS 14-26 via OS-level display-affinity APIs; responsibly disclosed. DOI: 10.5281/zenodo.20376495.
3. Aion-BibleQA: RAG citation-faithfulness & adversarial-robustness benchmark (40 questions); R@5 = 0.941, zero unsupported citations, 6/6 false-premise refusals. DOI: 10.5281/zenodo.20522874.
4. Project Zurvan: Local-first LLM knowledge engine extracting structured claims, concepts, entities, and decisions from documents, exposed to AI agents via an MCP stdio server. Python, SQLite FTS5 + embeddings, 218 tests.
5. Syllabus-Sync: Agent-readable campus-AI platform accepted into the Macquarie University startup incubator as a formal startup. Next.js 16, Supabase, WebAuthn passkeys, 503 tests across 92 files. Live at syllabus-sync.app.
6. Mehr Guard: Privacy-first offline phishing & QR/URL threat scanner. Kotlin Multiplatform (5 targets), ensemble ML + 25+ heuristics, 87% F1, sub-5ms latency, 1,248+ tests, and a 19-scenario built-in red-team suite.
7. SentinelFlow: Real-time C++ intrusion detection system parsing 500K+ packets/sec (single-thread parse) with signature and stateful detection of port scans, SYN floods, and DNS tunnelling.
8. NanoMatch: C++20 limit-order-book matching engine at 1M+ orders/sec (single-thread, synthetic benchmark) with O(1) cancellation.
9. eBPF Cloud Runtime Security Monitor (ECRSM): Educational read-only runtime visibility stack using kernel eBPF tracepoints, a Go agent, and a React dashboard.

## Professional Experience
- AI Safety Evaluator, Claude, Alignerr (Anthropic AI safety-evaluation program), Jan–Mar 2026.
- Freelance Security Engineer & Full-Stack Developer, Self-Employed, Jan 2024 – Present.
- IT Manager & Security Operations, Iran Pharmacy, 2019–2024.

## Philosophy
Raouf operates at the intersection of "Build" (scalable architecture) and "Secure" (defensive design), advocating privacy-by-design, local-first processing, and security embedded from design to deployment, now focused on measuring and containing the cyber capabilities of frontier AI.

## Contact
Website: https://raoufabedini.dev
Email (professional): raoof.r12@gmail.com
Security reports (RFC 9116): raoof.r999@outlook.com · https://raoufabedini.dev/.well-known/security.txt
GitHub: https://github.com/Raoof128
LinkedIn: https://linkedin.com/in/mohammad-raouf-abedini-885a9226a
X (Twitter): https://x.com/Raoofr12
ORCID: https://orcid.org/0009-0000-6214-258X

## Canonical Links (Markdown)
- [Homepage](https://raoufabedini.dev/)
- [Résumé](https://raoufabedini.dev/resume) · [Résumé PDF](https://raoufabedini.dev/Mohammad_Raouf_Abedini_Resume.pdf)
- [Projects](https://raoufabedini.dev/projects) · [Write-ups](https://raoufabedini.dev/write-ups) · [Lab](https://raoufabedini.dev/lab)
- [Contact](https://raoufabedini.dev/contact) · [Security Policy](https://raoufabedini.dev/security-policy) · [Hall of Fame](https://raoufabedini.dev/hall-of-fame)
- [GitHub](https://github.com/Raoof128) · [LinkedIn](https://linkedin.com/in/mohammad-raouf-abedini-885a9226a) · [ORCID](https://orcid.org/0009-0000-6214-258X)

### Flagship projects
- [Project Simurgh](https://raoufabedini.dev/projects/project-simurgh) · repo https://github.com/Raoof128/Project-Simurgh · DOI [10.5281/zenodo.20374849](https://doi.org/10.5281/zenodo.20374849)
- [The Invisible Window](https://raoufabedini.dev/projects/invisible-window-research) · DOI [10.5281/zenodo.20376495](https://doi.org/10.5281/zenodo.20376495)
- [Aion-BibleQA](https://raoufabedini.dev/projects/aion) · DOI [10.5281/zenodo.20522874](https://doi.org/10.5281/zenodo.20522874)
- [Project Zurvan](https://raoufabedini.dev/projects/project-zurvan)

### Feed
- [Write-ups RSS feed](https://raoufabedini.dev/feed.xml)