Principal AI/ML Engineer - AI Safety & Evaluation

About the Team 
We’re building a future where AI systems are not only powerful but safe, aligned, and robust against misuse. Our team focuses on advancing practical safety techniques for large language models (LLMs) and multimodal systems—ensuring these models remain aligned with human intent and resist attempts to produce harmful, toxic, or policy-violating content. 

We operate at the intersection of model development and real-world deployment, with a mission to build systems that can proactively detect and prevent jailbreaks, toxic behaviors, and other forms of misuse. Our work blends applied research, systems engineering, and evaluation design to ensure safety is built into our models at every layer. 

About the Role 

We’re looking for a Principal Engineer to lead the technical strategy and architecture for protecting foundation models against misuse—such as jailbreaks, prompt injection, toxic outputs, and custom policy violations. In this role, you’ll apply your expertise in scalable systems design, applied machine learning, and model-level defenses to build core infrastructure that ensures AI systems behave safely and responsibly in production. You’ll set technical direction and drive architectural decisions across a broad surface area of AI safety systems—designing safety interventions, integrating evaluation workflows, and developing models and tooling that detect and prevent harmful or non-compliant behavior. This role is ideal for someone who wants to work at the intersection of model behavior, product safety, and system engineering. 

What You’ll Do 

• Architect and lead the development of model-level defenses against jailbreaks, prompt injection, and custom policy violations 

• Define and drive evaluation strategies, including adversarial testing and stress-testing pipelines, to identify safety weaknesses before deployment 

• Set technical direction for scalable mitigation techniques such as safety-focused fine-tuning, prompt shielding, and post-processing methods to reduce harmful or non-compliant outputs 

• Collaborate with red teamers and researchers to convert emerging threats into measurable evaluations and system-level safeguards 

• Scale and improve human-in-the-loop pipelines for detecting toxic, biased, or non-compliant outputs 

• Stay up to date with LLM safety research, jailbreak tactics, and adversarial trends, and apply insights to real-world defenses  

What We’re Looking For 

• 7+ years of experience in applied machine learning, AI infrastructure, or safety-critical systems, with 3+ years in a senior or staff-level technical leadership role 

• Deep understanding of transformer-based architectures and experience building or evaluating safety interventions for LLMs 

• Proven expertise in analyzing and addressing adversarial behaviors, edge-case failures, and misuse scenarios 

• Demonstrated ability to guide long-term technical strategy, influence organizational direction, and mentor cross-functional teams 

• Strong written and verbal communication skills, with experience influencing technical direction at the org or platform level 

• Bachelor’s, Master’s, or PhD in Computer Science, Machine Learning, or a related field 

Nice to Have 

• Experience applying techniques such as reinforcement learning from human feedback (RLHF), adversarial training, or safety fine-tuning at scale 

• Hands-on work designing prompt-level defenses, content filtering systems, or mechanisms to prevent jailbreaks and policy violations 

• Contributions to AI safety research, industry standards, or open-source tools related to model robustness, alignment, or evaluation 

• Familiarity with model governance frameworks, including safety policies, model cards, red teaming protocols, or risk classification methodologies 

A10 Networks is an equal opportunity employer and a VEVRAA federal subcontractor. All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability status, protected veteran status, or any other characteristic protected by law. A10 also complies with all applicable state and local laws governing nondiscrimination in employment.

#LI-AN1 - Hybrid

Targeted compensation guideline: $225,000 - $245,000. Compensation will vary based on number of factors, including market demand for specific skills, role type, job level, and individual qualifications. Final salary offers are determined by considerations including, but not limited to, subject matter expertise, demonstrated skill level, relevant experience, geographic location, education, certifications, and training.