Parallel connects experienced cybersecurity specialists with projects supporting leading AI labs. We’re hiring a Vulnerability Researcher to help AI systems accurately distinguish between benign and malicious cybersecurity prompts.

You’ll apply deep expertise in vulnerability research, reverse engineering, and exploit development to evaluate complex, ambiguous prompts. Your work will involve determining true user intent, assessing technical context, and providing clear annotations that improve model safety, training, and evaluation guidelines.

This is remote, project-based work. Engagements and hours vary with project demand, with scope and expected availability discussed before each assignment.

What you’ll do

• Review ambiguous cybersecurity prompts submitted to AI systems to determine the underlying intent and goal.
• Assess whether prompts reflect legitimate research and defensive purposes or malicious attack preparation.
• Apply deep domain expertise to evaluate the technical context, specificity, and real-world uplift a response would provide an attacker.
• Annotate prompts with accurate intent labels and clear, concise written rationales.
• Collaborate with other researchers through independent dual-annotation, escalating disagreements to a reviewer.
• Identify and flag emerging attack techniques, edge cases, and gaps in current project guidelines to improve the evaluation process.

What you’ll bring

• At least five years of hands-on experience in vulnerability research and reverse engineering.
• Demonstrated experience discovering vulnerabilities using fuzzing, static and dynamic analysis, and patch diffing.
• Strong exploit development skills, including memory corruption, heap exploitation, ROP, and mitigation bypasses (ASLR, DEP, CFI).
• Fluency in decompilers (IDA, Ghidra, Binary Ninja) and debuggers (WinDbg, GDB, LLDB) across x86/x64 and ARM binaries.
• Deep understanding of the CVE lifecycle, n-day vs. 0-day research, and the weaponization of public PoCs.
• Strong analytical judgment and clear written communication in English to concisely document reasoning.
• Ability to work independently, manage details effectively, and solve complex, ambiguous problems.

Preferred experience

• Advanced offensive-security certifications such as OSED, OSEE, or GIAC GXPN.
• Demonstrated online achievements, published security research, or technical write-ups.
• Strong contributions on platforms like HackTheBox, Pwn2Own, TryHackMe, or Bug Bounty programs.

Selection process

Applicants must pass a practical assessment and complete a background check before engagement. Final compensation within the posted range will reflect relevant experience and demonstrated technical ability.