website · research (doi) · linkedin · pypi · npm · seotarek@gmail.com
I am a Principal Systems & AI Security Architect with 17+ years of experience designing high-reliability distributed platforms, deterministic AI agent runtimes, and memory-safe native systems in Rust, Python, Go, TypeScript, and Kotlin.
My research and engineering work centers on a strict architectural invariant: security enforcement, authorization boundaries, and clinical computation must be deterministic, never delegated to probabilistic model inference. Whether hardening Model Context Protocol (MCP) tool loops against indirect prompt injection (DOI: 10.5281/zenodo.22731104), scaling Rust graphics engines to billion-pixel canvases, or building regulatory-aligned clinical decision support, every guarantee is backed by formal constraints and executable verification suites.
- AI Agent Security & Runtime Firewalls: Sub-millisecond (
< 0.1 ms) defense against Indirect Prompt Injection (IPI), least-privilege MCP policy enforcement, and deterministic static/runtime gates. - Systems & Native Rust Engineering: Memory-safe rasterization at extreme canvas scale, binary format parsers, and privacy-first native desktop engines.
- Clinical Informatics & Regulated AI: Risk-stratified medical AI ecosystems aligned with IEC 62304 and ISO 14971, powered by zero-token deterministic clinical calculators and epistemic provenance locks.
- Distributed Infrastructure & BiDi Systems: High-concurrency Go messaging runtimes, PostgreSQL
pgvectorhybrid retrieval with AST query auditing, and W3C CSS Logical RTL/BiDi engineering.