Competed in Gray Swan's red-teaming arena, crafting indirect prompt-injection payloads to hijack tool-using AI assistants through poisoned web content.
Anthropic-backed red-teaming challenge focused on bypassing safety protocols and content filters in agentic AI systems to reach restricted objectives.
Engineered adversarial prompts to stress-test frontier AI models’ reasoning on production-grade security and cryptography problems, exposing flaws in authentication, authorization, and cryptographic design.
Evaluated model-generated code for injection vectors, broken crypto, authorization bypasses, and concurrency bugs, and authored production-quality reference solutions used as training data for subsequent model generations.
Competency-based program covering AI security, penetration testing, network security, cloud security, digital forensics, incident response, cryptography, and identity & access management.
Open to new opportunities. Feel free to reach out for collaborations or just to say hello.