You design and lead a multi-layered defense strategy against jailbreaks, prompt injection, data exfiltration, and tool misuse through advanced input/output scanners, safety filters, and autonomous agents. * You establish and continuously improve the security lifecycle for LLMs/Agents: threat modeling, attack simulations, red teaming, LLM-specific pentests, automated security assessments, and incident response frameworks. * Deep expertise in LLM and agent security: advanced protection against jailbreaks, prompt and indirect injection, input/output scanners, policy engines, and moderation strategies.
mehr