Staff / Senior Software Engineer (Agentic Search) - Runtime at Nebius

The Product

In a rapidly evolving world, trust in AI depends on AI agents being grounded in fresh, verified real-world data. Search is the foundation that makes this possible.

We are building an agent-native search platform designed specifically for AI systems rather than human users. Our product provides programmatic, low-latency, and observable search APIs that AI agents use to retrieve, filter, and reason over real-world information at scale.

The Role

We are looking for a Senior Software Engineer to work on the runtime systems of a novel search engine tailored for agentic AI consumption.

In this role, you will focus on building low-latency, high-throughput systems that serve search queries in real time. You will work on the critical path of user-facing requests, where performance, predictability, and efficiency directly impact product quality. You will design and operate systems that handle thousands of requests per second under strict latency budgets, optimising every layer from request handling to data access and response assembly.

In this position, your responsibility will be to

• Design, implement, and operate core runtime services for serving search queries at scale
• Build and optimise request flows, including query processing, retrieval orchestration, and response assembly under strict latency budgets
• Develop systems that maintain performance and predictability under high load
• Optimise CPU, memory, and data access patterns in performance-critical paths
• Ensure reliability, observability, and predictability across production services
• Build well-tested systems with clear responsibilities and interaction contracts, while remaining flexible as architecture evolves
• Define and implement observability primitives, including structured logs, metrics, traces, and latency breakdowns
• Monitor throughput, latency, and resource usage, and drive improvements in performance and cost efficiency
• Collaborate with indexing and ML teams to integrate retrieval and ranking components, keeping ML logic decoupled from core system internals
• Support experimentation and iteration through controlled rollouts and rigorous benchmarking

You may be a good fit