AV // SEC

Curriculum Vitae

Independent Systems & AI Security

[ 01 // Executive Profile ]

Security researcher and founding engineer focused on AI/ML systems, offensive security, and trust-boundary analysis. Specializing in mechanistic interpretability, model alignment, cgroups/eBPF sandboxing, and autonomous agent safety. Proven track record of high-severity coordinated vulnerability disclosures and formal systems research.

[ 02 // Professional Experience ]

Founding Engineer

SecNode
2025 — Present
  • Architecting autonomous offensive security execution pipelines and multi-agent workflows using LangGraph and model-routing abstractions.
  • Engineering container isolation boundaries using minimal cgroups cgroupv2 configurations and eBPF network socket filters to enforce rigorous sandboxing of untrusted tool execution.
  • Developing automated vulnerability auditing modules with custom parser analyzers to detect and flag trust-boundary escalation risks in runtime nodes.

Independent Security Researcher

AI & Distributed Systems Research
2024 — Present
  • Isolated safety-refusal representation subspaces in Qwen-3.5 residual stream layers; coordinate-disclosed low-rank abliteration techniques with zero capability degradation.
  • Discovered and patched CVE-2026-40281 (Critical CVSS 9.8 container socket escape / RCE) in autonomous agent orchestration frameworks via coordinated vulnerability disclosures.
  • Identified and securely reported high-severity XML Entity Expansion / Billion Laughs vulnerabilities to NASA's Vulnerability Disclosure Program via Bugcrowd.
[ 03 // Peer & Independent Publications ]

Technical analysis isolating safety-refusal pathways in the late transformer residual stream layers and projecting out refusal steering vectors via targeted weight orthogonalization.

Academic exploration modeling non-deterministic privilege escalation vectors, dynamic tool introspection abuses, and context-pollution exploits inside autonomous agent orchestrators.

[ 04 // Systems Engineering & Tooling ]

SecNode API Pentester

AI-augmented schema-driven microservices auditor focused on automated OpenAPI ingestion, multi-stage fuzzing pipelines, and authorization-bypass discovery.

Kai Agent Framework

High-performance Rust-based defensive scanner executing offensive vulnerability assessments inside strict eBPF containment boundaries.

PurgeProof

Cross-platform block-level data sanitization engine built to comply with NIST SP 800-88 Rev.1 secure erasure and auditable reporting standards.

[ 05 // Technical Skill Matrix ]
Models & InterpretabilityPyTorch, Transformer Lens, HookedTransformer, Hugging Face, Refusal Isolation & Abliteration
Offensive Systems SecurityContainer Sandboxing breakout vectors, eBPF socket filters, cgroups limits, threat modeling, XML parser attacks, API fuzzer design
Languages & RuntimesPython, Rust, Go (Golang), Linux, cgroupv2, gRPC, Docker, Git, SQLite / PostgreSQL
[ 06 // Academic Education ]
B.S. in Computer Science & Security EngineeringSpecialization in Distributed cgroups Containment and Deep Learning
Graduated // Independent Focus