Senior Staff Machine Learning Engineer

ServiceNow is an AI control tower for business reinvention, bringing together AI, data, and workflows to help enterprises work smarter, faster, and better. The Senior Staff Machine Learning Engineer will design, build, and scale production-grade agentic AI systems, including multi-agent architectures, enterprise-grounded reasoning, retrieval, model integration, safety, governance, and robust infrastructure. The role also provides technical leadership for architectural decisions, design reviews, and production AI practices.

Responsibilities:

• Agentic architecture. Design and ship multi-agent systems — orchestration, tool use, planning loops, memory, and failure recovery — that operate reliably in production, not in notebooks
• Enterprise-grounded reasoning. Build agents that leverage ServiceNow's data layer — CMDB, Workflow Data Fabric, and Knowledge Graph — to make decisions with context no frontier model has on its own
• Trust, safety, and governance. Own the guardrails: observability, human-in-the-loop controls, and compliance infrastructure that make autonomous systems safe to deploy at scale
• Retrieval and grounding. Work closely with our search team to ensure agents are grounded in accurate, low-latency retrieval — RAG pipelines, hybrid search, re-ranking, and evaluation — as a critical dependency of agentic quality
• Model integration and evaluation. Integrate frontier models (Anthropic, Google, OpenAI) into the Sense → Decide → Act → Govern architecture; evaluate trade-offs across cost, latency, and capability for production use cases
• Technical leadership and strong bias for action. Set the architectural patterns the group works from. Own the hard design calls, run the design reviews, and raise the bar on agentic design and production AI discipline across engineers and principals
• Designing scalable and robust architectures that will support at scale deployment across hyperscalers and our own infrastructure
• Work on emerging model capabilities and applying them to real world customer problems on a short timeline