AI Agents for Security

My work on security agents covers two areas: using agents to assess threats and developing controls on the tools and resources agents can access.

Multimodal phishing detection

Role: ML Research Engineer, Huawei Canada

I designed a multimodal phishing detection agent that coordinates LLM calls for evidence-based threat assessment.

The design connects three parts of the assessment workflow:

• Evidence collection: Extract relevant information from the email, its attachments, and its URLs.
• LLM orchestration: Coordinate model calls using the collected evidence.
• Threat assessment: Use the evidence and model responses to assess phishing threats.

Agent permission controls

Role: Senior ML Research Engineer, Huawei Canada

I develop permission controls for AI agents to govern tool access and resource use across agentic workflows.