About the job
We are seeking a Trustworthy AI, Safety Architect to lead safe AI innovation across global markets. This is a hybrid role blending AI governance, technical safety, hardware-based trust, and strategic ecosystem engagement. You will shape how governments, enterprises, and partners deploy open, safe, and sovereign AI on AMD solutions, across AMD’s broad AI Infrastructure portfolio (incl. GPU, CPU, AI PCs, FPGAs). Ensuring Responsible and Trustworthy AI from cloud-scale systems all the way to the edge.
Responsibilities
Lead global Trustworthy AI strategy aligned with ISO/IEC 42001, NIST RMF, EU AI Act.
Design and implement AI safety guardrails, dialogue governance, jailbreak defense, content moderation, and accuracy assurance pipelines.
Drive program management for safe-by-design development across hardware, software, and model teams.
Translate silicon-level trust features—attestation, secure enclaves, TPM, measured boot, watermarking, provenance metadata—into differentiated safety solutions for global partners.
Define “trusted inference mode” by integrating hardware signals with software guardrails.
Build research and policy collaborations with universities, governments, AAAI, standards bodies, NGOs, and industry alliances.
Represent the organization in global forums, conferences, and executive briefings.
Architect safe AI deployment patterns for national AI strategies, enterprise rollouts, and sovereign AI initiatives.
Qualifications
Minimum
No minimum qualifications listed.
Preferred
Expertise in SLMs/LLMs, transformer architectures, inference optimization, and on-device AI.
Strong engineering background in Python, C/C++, or Rust.
Experience with safety tuning, RLHF/RLAIF, policy enforcement layers, and red-teaming.
Understanding of provenance, trusted compute, and security architectures.
Strong command of AI risk frameworks and global regulatory landscape.
Ability to convert governance requirements into technical controls and partner-ready solutions.
Demonstrated cross-functional leadership across hardware, software, policy, and research teams.