Senior Staff ML Engineer, AI Safeguards - Moveworks
Remote • Mountain View, CALIFORNIA, United States • Full-time
Posted 15h ago
Job Location
Mountain View, CALIFORNIA, United States
Tech Stack
Remote Work Policy
Fully remote
Employment Type
Full-time
Categories
Machine Learning Engineer
About the job
Moveworks is seeking a Senior Staff ML Engineer to join our AI Safeguards team. In this role, you will be instrumental in designing, building, and shipping critical safeguards within our agent harness. Your work will focus on protecting against prompt injection, single and multi-turn jailbreaks, toxicity, sensitive topics, and sensitive data exposure. You will develop robust evaluation frameworks to measure the efficacy of these safeguards, addressing both false positives and false negatives using diverse datasets. Collaboration with our Security team will be key, translating their findings into actionable eval cases and fixes, and co-developing mitigations. You will also select, integrate, and fine-tune detection models within a shared stack serving both commercial and government environments, while carefully balancing quality, latency, and cost for controls operating in the request path. Additionally, you will build configurable safeguard policies to empower customers in managing sensitive topics and monitor safeguard behavior in production, addressing any regressions.
Responsibilities
- Design, build, and ship safeguards for prompt injection, jailbreaks, toxicity, sensitive topics, and sensitive data exposure.
- Build evaluations to measure safeguard efficacy using various datasets.
- Collaborate with the Security team to translate adversarial findings into eval cases and fixes.
- Select, integrate, and fine-tune detection models for production environments.
- Balance quality, latency, and cost for request path controls.
- Develop configurable safeguard policies for customer customization.
- Monitor safeguard behavior in production and fix regressions.
Requirements
- 8+ years of experience building production software.
- 3+ years of applied ML experience (training, evaluating, and deploying models).
- Hands-on experience developing LLM safeguards (e.g., guardrail models, content classifiers, prompt injection/jailbreak detection, safety evals).
- Strong model evaluation fundamentals for agent trajectories and text classification.
- Understanding of LLM safeguard bypass techniques (e.g., indirect prompt injection, multi-turn jailbreaks).
- Excellent Python experience.