Whitepapers.
Peer-style technical papers authored by AFSP-credentialed engineers from active safety programs. Distinct from field notes: numbered sections, figures, tables, and references.
Evidence Generation for Neural Networks in Safety-Critical Embedded Systems
Certifying neural networks under IEC 62304 Amendment 1 and the EU AI Act simultaneously requires a structured evidence generation strategy that satisfies both frameworks without doubling engineering effort. This paper presents a dual-compliance methodology covering dataset governance, V&V strategy, coverage criteria, and technical documentation, with worked examples from medical imaging and industrial quality inspection.
Safety Assurance Framework for Physical AI Systems
Physical AI systems — autonomous vehicles, collaborative robots, surgical platforms — must satisfy safety certification regimes built for deterministic embedded software. This paper proposes a unified assurance framework that maps ISO/PAS 8800:2025, UL 4600:2023, and ISO 21448 SOTIF onto a shared evidence structure, with worked examples in L3 highway automation and a surgical robotic arm.
Systematic ASIL Decomposition in Multi-Channel E/E Architectures
ASIL decomposition allows a single high-integrity requirement to be split across two independent channels, each carrying a lower ASIL rating. This paper presents a structured methodology for applying Part 9 of ISO 26262:2018, with worked examples in brake-by-wire and EPS architectures, hardware metric verification, and dependent-failure analysis.