CybersecurityJobs.io
← Back to all jobs

Job Description

Gracemark is hiring a Cybersecurity SME to strengthen AI safety through hands-on offensive evaluation. This hybrid role in Southlake, TX focuses on probing AI models and AI coding agents in sandboxed environments to assess real-world cyber risk. If you value meticulous testing, clear documentation, and mentoring alongside emerging AI attack research, this role blends offensive security with structured safety evaluation.

What you’ll do

  • Iteratively probe AI coding agents and frontier LLMs in sandboxed environments to determine whether models can produce functional offensive tooling, exploits, and malware.
  • Assess model outputs against established safety specifications, gauge severity of any violations, and refine evaluation criteria for unsafe cyber capability uplift.
  • Evaluate performance across vulnerability discovery, exploit development, and malware analysis, including verification that AI-generated tools actually execute as intended within sandboxes.
  • Investigate agent-specific attack surfaces such as prompt injection, tool misuse, and privilege escalation, especially as assessments expand into more complex agent harnesses.
  • Build realistic attack scenarios and evaluation datasets, and review expert-written prompts and model outputs for technical accuracy.
  • Document reproduction steps, severity ratings, and mitigations; map findings to MITRE ATLAS, OWASP Top 10 for LLM Applications, and NIST AI RMF.
  • Train and mentor red teamers and annotators on offensive security concepts while staying current on emerging AI attack techniques.

What you bring

  • 2+ years of hands-on offensive security experience, including penetration testing, red teaming, vulnerability research, or exploit development.
  • Direct experience with exploit development, malware behavior, post-exploitation techniques, and web or cloud attacks.
  • Strong scripting skills (preferably Python) with the ability to read, debug, and safely verify AI-generated offensive code in a sandbox.
  • Working knowledge of LLMs, AI coding agents, and common attack vectors targeting them (such as jailbreaks, prompt injection, and tool misuse).
  • Strong written communication to document reproduction steps and explain complex security risks to researchers, engineers, and executives.
  • Ability to handle sensitive or adversarial cyber content responsibly while following confidentiality and information-handling protocols.

Helpful background

  • Certifications or a proven track record via OSCP, OSEP, GPEN, published CVEs, CTF placements, bug bounties, AI/ML security research, defense/intelligence background, malware reverse engineering, or cloud security (AWS, GCP, Azure).

Tools and focus areas

  • Python, LLMs, AI coding agents
  • MITRE ATLAS, OWASP Top 10 for LLM Applications, NIST AI RMF
  • AWS, GCP, Azure

Compensation and location

Pay: $65.00 - $70.00 per hour

Expected hours: 40.0 per week

Location: Hybrid remote in Southlake, TX 76092

Benefits

  • 401(k)
  • Dental insurance
  • Health insurance
  • Vision insurance

Similar Jobs