Biró István
Building things. Writing about them.
Requirements Engineer at Hirschmann Car Communication and founder of BWEngineering — an engineering and additive manufacturing studio. I work across systems architecture, hardware-software integration, and the design of agentic AI workflows. Featured in SZE Admissions.
What I work on
- Systems Engineering
- Full V-model lifecycle work — requirements through architecture, functional decomposition, and verification and validation. My background covers automotive E/E systems at Bosch (Electronic Battery Sensor) and Jaguar Land Rover (ADAS, Surround View): MBSE with Magic Systems of Systems Architect, requirements management in IBM DOORS and 3DEXPERIENCE TRM, and systems test across multiple OEM platforms. Joining Hirschmann Car Communication as a Requirements Engineer in July 2026.
- Engineering Studio
- BWEngineering is the engineering and additive manufacturing studio I run alongside the day job. Projects span mechanical design, FEA and simulation, and small-series production — from concept to physical artifact. Based in Hungary, serving clients across Central Europe on commissions that range from structural analysis to functional prototyping.
- Agentic AI Workflows
- I design and operate human-governed AI workflows: decomposing complex work into agent-ready specifications, maintaining audit trails, and building systems where humans own decisions and machines own implementation. This portfolio site is a working demonstration — every decision is documented, every line of code was written by an implementation agent from committed specifications, and every merge is governor-approved.
- Teaching & Research
- During my time at Széchenyi István University I taught Technical Drawing and Machine Design, mentored undergraduate students, and conducted research in applied mechanics and vehicle dynamics. Two peer-reviewed publications came from that period. The academic habit of documenting decisions rigorously has stayed.
From the blog
-
The AI Illusion: The Model Isn't Autonomous. The System Is
Why corporate AI fears focus on the wrong thing: the model is a stored mathematical object, and the real security boundary is the context, memory, tools and permissions built around it.
-
The AI Sandbox Problem: Trust Boundaries, Tool Execution and Enterprise Security
A Microsoft Copilot bug bounty report closed as "not a vulnerability" revealed something more interesting: the gap between what an AI model refuses to do and what its execution sandbox is actually capable of.