Projects

A multi-agent CTF harness built with Squid Proxy Lovers, combining category-specialized research agents with reproducible evaluation.

• ai-security
• ctf
• evaluation

• Built category-specialized agents and integrated containerized execution, retrieval, handoffs, and result validation.
• Team submission: 44 of 50 challenges solved at CSAW Agentic Automated CTF.
• Subsequent development: 46 of 50 challenges solved on NYU CTFTiny (92%).

A resumable security-analysis platform for organizing code review, triage evidence, evaluation, and reporting.

• ai-security
• static-analysis
• tooling

• Built a 19-stage audit workflow with checkpoints and session logs.
• Combined static scanning, file profiling, and program-structure analysis to support evidence-based review.
• Integrated asynchronous Semgrep verification and evaluation across CVE datasets.