Adversarial AI Research Scientist

Peraton Labs is looking for a self-motivated Adversarial AI Research Scientist with a passion for developing solutions for adversarial AI/ML (Artificial Intelligence/Machine Learning). Recent research has shown that AI-based systems are highly prone to error when presented with adversarially crafted inputs, such as imperceptible perturbations added to an image. Other sources of error arise from adversarial manipulation of training data. Such adversarial actions can be used to disrupt the performance of AI-enabled physical systems, such as autonomous ground vehicles and unmanned aerial vehicles. Such disruptions result in degraded performance (e.g., autonomous vehicles slowdown) or failure to accomplish a task (e.g., vehicle crashes). However, most of these effects have been demonstrated in simulation environments or assume unrestricted access to the model training pipeline, which is unrealistic. Peraton Labs is a leader in the field of Adversarial AI and we are seeking strong candidates to join our team and investigate challenging new problems in this space.

You will work with a small team of research scientists to investigate the vulnerabilities of AI-enabled systems in real-world settings to attacks across multiple modalities and at different stages of the model training pipeline. You will devise physically realizable solutions to determine whether real-world systems like autonomous vehicles and unmanned aerial vehicles are vulnerable to attacks on their AI controls. Note that such vehicles will depend on sensors that span multiple modalities, such as visual, thermal, LiDAR, etc.; thus, for attacks to succeed, each of these modalities must be manipulated. Your work will involve research, design, and prototyping of software as well as field-testing to prove out technical concepts; you should therefore have strong software development skills and be willing to learn new skills and technologies as needed to develop cutting-edge solutions based on customer needs.

The ideal candidate is one who is excited at the opportunity to work on cutting-edge research in an environment where they will continuously learn new concepts and innovate in collaboration with a strong team of individuals with diverse technical backgrounds.

Required Qualifications

• A minimum of a MS with 3 years of experience, or a Ph.D. in Computer Science, Computer Engineering, or related field with a focus on machine learning research.
• Hands-on experience with state-of-the-art machine learning software environments such as TensorFlow, PyTorch, scikit-learn.
• Strong software development experience.
• Excellent oral and written communication skills.
• US citizenship is required for this position.
• Candidates must be local to or willing to relocate to the immediate Basking Ridge, NJ area.

Preferred Qualifications

• A strong publication record in AI/ML.
• Expertise and research experience specific to adversarial machine learning (AML) is desirable.