SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Netskope is a cloud security company founded in 2012, redefining how organizations protect data and networks as the traditional perimeter dissolves. The company operates globally with offices across Santa Clara, St. Louis, Bangalore, London, Paris, Melbourne, Taipei, and Tokyo.
As a Senior Agent Runtime Engineer, you will own the health and reliability of the execution environment where agents run in isolation. Your core responsibilities include:
• Maintain session isolation and resource limits across the agent platform, identifying failure modes that only surface under real concurrency and production traffic patterns rather than in controlled test environments.
• Investigate and resolve production discrepancies where agents behave differently than expected, collaborating directly with agent and quality engineers to determine whether issues stem from agent design or environmental factors.
• Tune cold-start and concurrency settings for critical-path functions, regularly reviewing and adjusting configurations as usage patterns evolve rather than applying static settings.
• Deeply understand the platform's session isolation model, knowing where guarantees come from the underlying compute layer versus where the platform must implement additional controls.
• Build and maintain comprehensive observability—tracing, behavioral drift detection, and quality signals—that answers whether agents are working correctly, not just whether they're technically operational.
• Monitor per-agent and per-session cost and efficiency metrics, flagging agents consuming excessive tokens, making unnecessary tool calls, or running longer than tasks should require.
• Deploy and manage runtime infrastructure resources through CI/CD pipelines, including provisioned concurrency settings, runtime-specific IAM roles, and observability configurations.
• Design and execute tests validating that isolation guarantees actually hold in practice, confirming agents cannot access or affect each other's sessions.
You'll be the person who understands complex systems deeply enough to diagnose exactly why they misbehave—the first call when something goes wrong in production but not in testing.
