Skip to main content

/Resume

Download Resume

Mohammad Raouf Abedini

AI Security Research · LLM Agent Red-Teaming · Offensive & Defensive Security Engineering

Sydney, Australia · Ready to relocate now to San Francisco, CA · Travel-ready (Washington, DC) · Visa sponsorship required
01.

About

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 from 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). Positioned as the defense-in-depth layer complementary to inline classifiers: they govern what a model may say; this attests what an agent was allowed to do. 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.

02.

Security Research

Node.js · Ed25519 attestation · Lean 4 · AgentDojo · Llama Guard 4 · Llama-3.3-70B · JS↔Python parity

Project Simurgh — Verifiable Containment Attestation for Agentic AI (Creator · AGPL-3.0)

  • Built a provider-agnostic framework producing Ed25519-signed, offline-reproducible evidence of what an AI agent did after a guardrail miss, across four containment boundaries: input firewall, context-provenance guard, tool-invocation gate, and output-leakage firewall — under an adversarial, dishonest-producer threat model
  • Guardrail-miss containment: captured a real Llama Guard 4 (12B) input classifier over a 180-case run-set and contained 138/138 malicious cases the classifier missed (120 downstream-injection cases an input-only classifier structurally cannot see, 18 direct-input misses); combined targeted attack-success 0/150, with zero unsafe tool executions or exports
  • Live-agent containment: drove a self-hosted Llama-3.3-70B through AgentDojo's workspace suite (140 pre-registered injection cases); the tool-authority gate cut targeted attack success from 9/140 to 0/140 with benign utility held
  • Positioned as the defense-in-depth layer complementary to Anthropic's inline classifier — the classifier governs what a model may say, the attestation governs what an agent was allowed to do (non-claims signed and explicit, including that it would not have caught the June 2026 content-generation bypass itself)
  • Attacked its own proof (red-team sweep across eight attack classes → detector-v2) and formally checked oversight with five machine-checked Lean theorems; one-command offline reproduction of a 12-rung signed release ladder; 3,057 automated tests
03.

Technical Proficiencies

> Languages

Python (primary), C, C++, TypeScript, JavaScript, Swift, Kotlin, Bash, SQL, Go (familiar)

> Security & Offensive

Vulnerability research, cross-platform exploit development (Win32 API, macOS ScreenCaptureKit), threat modelling, secure code review, penetration testing, responsible disclosure (OWASP/FIRST/CISA), Wireshark, Nmap, Burp Suite

> AI & ML

Large Language Model (LLM) integration & evaluation, Retrieval-Augmented Generation (RAG) evaluation, citation-faithfulness benchmarking, AI-assisted vulnerability research, Natural Language Processing (NLP), generative AI tooling, ML model evaluation, dual-use risk assessment

> Systems & Tools

Linux (Ubuntu/Kali), CMake, Docker, Git/GitHub, GitHub Actions CI/CD, Google Test, FastAPI, Cloudflare Workers, libpcap

> Frameworks

Open Web Application Security Project (OWASP) Top 10, MITRE ATT&CK, National Institute of Standards and Technology (NIST) Framework, W3C Screen Capture Specification

04.

Education

Bachelor of Cyber Security

Macquarie University
May 2024 – Nov 2026
Coursework: Digital Forensics, Network Security, Systems Security, Cloud Computing, Natural Language Processing (NLP) & Machine Learning, Privacy-Preserving Data Analysis

Diploma of Information Technology

Macquarie University
Jul 2023 – May 2024
05.

Selected Research & Engineering Projects

The Invisible Window [DISCLOSURE]

2026

C · Swift · Python · Win32 API · ScreenCaptureKit · WebRTC

IEEE-format vulnerability research: independently discovered and responsibly disclosed a cross-platform screen-capture evasion class exploiting OS-level display-affinity APIs (Windows/macOS) that defeats browser-based capture and AI-vision pipelines — 100% evasion across all tested platforms with zero visual artefacts over 10,000+ analysed frames, coordinated disclosure to OS and proctoring vendors. DOI: 10.5281/zenodo.20376495.

Aion [BIBLE RAG]

2026

React Native · Expo · Supabase · pgvector · Gemini · OpenAI Embeddings · Tauri v2

Built AI-powered Bible companion and authored Aion-BibleQA, an 8-page preprint introducing a 40-question benchmark for citation faithfulness and false-premise robustness — R@5 = 0.941, mean citation_support = 0.978, zero unsupported citations, and 6/6 false-premise refusals. DOI: 10.5281/zenodo.20522874.

NanoMatch [SYSTEMS]

2026

C++20 · CMake · Google Test

Engineered high-performance matching engine processing 1M+ orders/second with sub-microsecond latency — implemented red-black tree price levels, custom memory pool allocator, and comprehensive test suite with p50/p99 latency benchmarks.

SentinelFlow [IDS]

2026

C++17 · libpcap · CMake · Google Test · Linux

Built real-time network packet processing engine parsing 500K+ packets/second — protocol dissection (Ethernet/IPv4/TCP/UDP/ICMP/DNS), signature-based detection engine, and stateful analysis (port scans, SYN floods).

Nexus Archive [FULL-STACK]

2025

Python/Litestar · React · PostgreSQL · Docker · Terraform

Shipped full-stack data platform with AI recommendation engine, event-driven API design, rate limiting, and automated security scanning — end-to-end ownership from database schema to deployment infrastructure.

Mehr Guard [KOTLINCONF]

2024

Kotlin Multiplatform · Local ML · Android & iOS

Built cross-platform offline threat detection tool with local ML-based classification — submitted to KotlinConf global developer conference.

70+ additional public projects on GitHub covering vulnerability research, systems programming, AI/ML tooling, and cloud infrastructure: github.com/Raoof128
06.

Professional Experience

Freelance Security Engineer & Full-Stack Developer

Self-Employed · Jan 2024 – Present
  • Architected security backends for production, multi-frontend platforms on Supabase: WebAuthn passkeys, TOTP/SMS MFA, zero-trust edge middleware, distributed rate limiting, strict Postgres Row-Level Security, and tamper-resistant audit logging via SECURITY DEFINER RPCs
  • Engineered CI/CD with 500+ automated security tests (GitHub Actions), cutting deployment vulnerabilities ~40%; shipped security-first apps for 1,000+ users with OWASP Top 10 controls and OAuth 2.0
  • Sustained a 5.0-star rating across 31 client reviews and 35 completed tasks (95% completion, ID-verified) on Airtasker, delivering web development, Python, API integration, and IT/security work for paying clients

IT Manager

Iran Pharmacy · Aug 2019 – May 2024
  • Managed technology infrastructure across a multi-site organisation for 5 years — maintaining 99% system uptime, enforcing role-based access control (RBAC), and automating operational workflows via Python/Bash scripting (~30% reduction in manual tasks)
07.

AI Safety, Leadership & Community

  • AI Safety Evaluator, Claude (Alignerr, Anthropic AI safety-evaluation program, 2026): assessed Claude outputs for exploitable code and analysed how safety guardrails can be circumvented, delivering structured, rubric-based findings (continues earlier 2024 Claude Code evaluation work)

Co-Founder · Macquarie Persian Students Society

2026 – Present
Volunteering · Education

Founded and help run the university's Persian student community — events, peer support, and cultural programming.

08.

Licenses & Certifications

> Anthropic (10)

AI Fluency for Students

Issued Jul 2026 · Credential ID: acmpujtbn2xu

AI Fluency for Educators

Issued Jul 2026 · Credential ID: 7x8msbzn49rt

Introduction to Model Context Protocol

Issued Jul 2026 · Credential ID: yhi68u4mqt5x

Building with the Claude API

Issued Jul 2026 · Credential ID: juoae2qtmggo

AI Fluency: Framework & Foundations

Issued Jul 2026 · Credential ID: 645j7by2uo75

Claude Code in Action

Issued Jul 2026 · Credential ID: 7kfqpyogooec

Introduction to Claude Cowork

Issued Jul 2026 · Credential ID: bsqfvrbkpv2s

Claude Platform 101

Issued Jul 2026 · Credential ID: q8dcrer7o2pa

Claude Code 101

Issued Jul 2026 · Credential ID: oyf4ic48wnd5

Claude 101

Issued Jul 2026 · Credential ID: hyagrxaidsbe

> Macquarie University (13)

Cyber Security: GRC Part 2 — Risk Management and Compliance

99.20%

Issued Jun 2026 · Credential ID: JL4SAS9JOR8C

Cyber Security: GRC Part 1 — Governance

95%

Issued Jun 2026 · Credential ID: 3G4JP3GKROWQ

Cyber Security: Mobile Security

99.28%

Issued Jun 2026 · Credential ID: W7FRQ19BPTNM

Cyber Security: Applied Cryptography

98.56%

Issued Jun 2026 · Credential ID: OVVCSLEB8ATT

Cyber Security: Data Security and Information Privacy

97.84%

Issued Jun 2026 · Credential ID: K54QIRP9VVE9

Cyber Security: Digital Forensics

95%

Issued Jun 2026 · Credential ID: YX5P82YME6FQ

Cyber Security: Identity Access Management and Authentication

100%

Issued Jun 2026 · Credential ID: P1RJQUGSCM59

Cyber Security: DevSecOps

99.10%

Issued Jun 2026 · Credential ID: GJEBDPIA0A6P

Cyber Security: Application of AI

99.20%

Issued Jun 2026 · Credential ID: U2KSU5H0O0BV

Cyber Security: Security of AI

95.50%

Issued May 2026 · Credential ID: DNZVZ3A7RVR1

Cyber Security: Essentials for Managers and Leaders

99.40%

Issued May 2026 · Credential ID: LOBWM8KXGSZS

Cyber Security: Essentials for Workplace

92.80%

Issued May 2026 · Credential ID: QT8ZBJIJ2HHO

Cyber Security: Essentials

96%

Issued May 2026 · Credential ID: RNALWEYYSXO7

09.

Additional Information

Ready to relocate now to San Francisco; available to start immediately (final semester completed online). Visa sponsorship required.
English (Professional Working) · Persian / Farsi (Native) · Japanese (Elementary)