Senior Researcher, Agentic AI Security & Vulnerability Discovery

Hawksbill is dedicated to utilizing Artificial Intelligence to foster a more cyber-secure and ethically responsible world. The Senior Researcher will work at the intersection of ML engineering, security research, and systems design, focusing on building systems that learn to attack and contribute to innovative security solutions.

Responsibilities:

• Design and curate training datasets for vulnerability discovery — including synthetic generation, expert elicitation from security researchers, and structured capture of offensive reasoning traces
• Develop reward models and RLHF pipelines that shape model behavior toward accurate, novel, and reproducible security findings rather than surface-level pattern matching
• Build MCP servers that expose security-relevant capabilities to agentic models: static analysis tools, dynamic execution environments, CVE databases, decompilers, network interfaces, and custom exploit scaffolding
• Design and run evaluations that measure real security capability — CTF-style benchmarks, live code auditing tasks, and controlled zero-day hunting exercises against isolated targets
• Instrument agentic pipelines to capture failure modes: hallucinated findings, reasoning shortcuts, and unsafe behaviors when operating in adversarial environments
• Collaborate with red teamers to convert manual exploit techniques into automated training signal, closing the loop between human expertise and model capability
• Publish work externally — training methodologies, benchmark results, and novel findings — contributing to the emerging field of AI-native security tooling