Decagon
Research Engineer, Safety
About this role
Decagon seeks a Research Engineer focused on Safety to build safeguards and evaluations for conversational AI agents in enterprise deployments. You'll identify failure modes, develop classifiers and runtime protections against prompt injection and unsafe behaviors, and work across teams to scale safety practices from research to production.
What you'll do
- Research and implement safeguards against prompt injection, unsafe tool use, data disclosure, and policy violations
- Build adversarial evaluations, red-team datasets, and regression suites based on production failures
- Develop classifiers, reward signals, and post-training methods to improve agent safety
- Analyze production incidents to identify root causes and measure mitigation impact
- Partner with Security, Product, Infrastructure, and Legal teams to translate enterprise safety requirements into deployable systems
- Own technical decisions end-to-end from evaluation design through production rollout
What they're looking for
- Python and modern ML tooling
- Language model evaluation and post-training (RL, preference optimization, distillation)
- Adversarial testing and model red teaming
- Prompt injection and policy enforcement techniques
- Agentic systems design and deployment
- Production ML systems engineering
- Privacy and safe tool use
- Experimental design and technical judgment
Benefits
- Equity compensation in addition to base salary
- Work on cutting-edge applied AI safety in production
- High-impact technical ownership and autonomy
- Collaboration with world-class research and engineering teams
- In-office environment in San Francisco
Opens the application — the Jobs AI extension fills it for you. Set up autofill
Opens the official application on the employer’s site. No login required.
Decagon
Decagon builds enterprise-grade conversational AI platforms that enable organizations to deploy AI agents for business impact. The company is hiring Strategic Solutions Engineers, Customer Engineers, Platform Engineers, and systems-focused engineers to deliver AI implementations, build internal infrastructure, and establish security practices across their growing platform.
View all jobs at DecagonLikely interview questions
- Walk us through a time you identified a safety failure in a production ML system and how you approached fixing it.
- Describe your experience building or evaluating safeguards against prompt injection or adversarial attacks on language models.