AI Red Team Engineer

Who We Are

SEI conducts research and development in software engineering, systems engineering, cybersecurity, and many other areas of computing, working to introduce private-sector innovations into government. The SEI works closely with defense and government organizations, industry, and academia to continually improve software-intensive systems. Its core purposes are to help organizations improve software engineering capabilities, advance cybersecurity methods and technologies, and bring the discipline of software engineering to AI systems.

What We Do

The CERT Threat Analysis (TA) Directorate conducts research and development activities to identify, analyze, coordinate disclosure, and mitigate threats and vulnerabilities in systems and software. The TA Directorate is currently comprised of three teams: Artificial Intelligence (AI) Security, Malware and Vulnerability Exploitation, and Platform and Mission Engineering. The AI Security team works on advancing the state of the art in AI security at a national and global scale. The Malware and Vulnerability Exploitation (MVE) team works to improve cyber-tradecraft analysis within strategic target communities to counter adversarial use of the Internet and related technologies. The vulnerability side of MVE (home of the CERT Coordination Center) works with an expansive network of vendors, partners, and collaborators to reduce the societal harm of vulnerable software and systems. The Platform and Mission Engineering team develops and maintains tools, environments, and operational support for the malware analysis, reverse engineering, vulnerability analysis, and AI security domains.

Position Summary

As an AI Red Team Engineer on the AI Security team, you will play a central role in adversary emulation exercises and capability development for our mission partners. Due to our unique position within the TA Directorate, the systems we red-team fall outside the realm of 'traditional' enterprise red teaming. Our targets are commonly AI-enabled platforms used within national security contexts.

But this isn't a "make the LLM say the bad thing" type of AI red team. We operate across multiple domains, meaning that our red teamers are expected to be experts in offensive cyber in addition to AI security. If you are experienced with offensive cyber tradecraft and have an interest in breaking into AI, this could be a good fit. Most of our red teamers are actively taking graduate-level technical courses at CMU and/or pursuing technical certifications. Perpetual learning is a core part of what we do.

While our red team exists within a research organization, research is only a portion of the work performed by our red team. Much of the work will involve red teaming real-world systems, sometimes at an aggressive cadence. This can involve planning and rehearsing red team TTPs, traveling to field sites, and presenting relevant findings. Like most red teams, we don't get to pick and choose our targets. This means that our red team needs to be well-rounded (both as individuals and as a team). Thus, we expect all applicants to be savvy with both Windows and Linux, solid with TCP/IP, and have some experience with penetration testing and/or red teaming.

What you’ll do:

• Red team real-world AI-enabled systems (both the model and the hardware/software/network that it runs on) in support of national security objectives.

• Develop new tactics, techniques, and procedures for attacking AI-enabled systems and related software in order to better prepare defenders for real-world threats.

• Write tools in Python, PowerShell, C, and BASH to enable red team operations.

• Represent the CERT technical portfolio of work and operations; communicate with external mission partners and internal collaborators in concert with CERT directorates and teams.

Who you are:

• BS in computer science, software engineering, networking, information systems, or a related technical field with eight (8) years of experience; MS in computer science or technical/engineering field with five (5) years of experience; PhD in computer science or technical/engineering field with two (2) years of experience or equivalent combination of training and experience. Other educational backgrounds of a technical nature with experience as described may be considered.

• You have previous penetration testing, red teaming, or exploit development experience.

• You have previous hands-on experience with at least one command and control framework (e.g., Cobalt Strike, Sliver).

• You have experience programming/scripting in Python, C, and BASH (without the assistance of AI) and are willing to learn PowerShell.

• You have experience with reverse engineering tools (e.g. NSA Ghidra, IDA Pro).

• You are able to read code and quickly spot basic vulnerabilities without the assistance of AI or fuzzing.

• You are very familiar with TCP/IP and all layers of the OSI model. You have experience using Wireshark and can explain how common network protocols work.

• You have experience in assessing the security of both Linux and Windows systems. Experience with mobile (e.g., Android) and other operations systems is also appreciated.

• You have at least two of the following relevant certifications: OSCP, CPTS, FORGE/RIOT, GXPN, GAWN, GCPN, CRTO, CRTL, OSEP, OSWE, CCNA, CWEE. Applicants without these certifications will still be considered if equivalent experience is clearly demonstrated during technical interviews.

• You have a willingness to travel (25%) outside of your office location to other SEI offices, sponsor sites, conferences, and offsite meetings.

• You have excellent communication skills (oral and written), particularly regarding technical communications with non-experts.

• You enjoy mentoring and cross-training others and sharing knowledge within the broader community.

• You will be subject to a background investigation, and you must have the ability to obtain and maintain a Department of War security clearance.