Member of Technical Staff (Site Reliability Engineer, Crawler)

Keeping a crawler running is not the same as keeping it productive. Queues can grow while servers look healthy, and a small change can become a fleet-wide bottleneck. On Perplexity's Crawler team, you'll improve throughput, plan capacity, and build operational reliability through safer deployments, better observability, and durable fixes. Responsibilities - Operate and optimize Kubernetes workloads and supporting infrastructure for crawler services, including fetching and rendering fleets. - Build and maintain infrastructure as code and CI/CD pipelines, with safe rollouts, validation, and fast rollback. - Define SLOs and build monitoring, alerting, and observability for availability, latency, throughput, and resource saturation. - Own capacity planning and resource efficiency; diagnose queue backlogs, network failures, and resource bottlenecks with service owners. - Participate in the team's on-call rotation, lead incident response, and turn blameless postmortems into durable reliability improvements. - Write production software to automate diagnosis and routine remediation, with validation and approval gates for high-impact actions. Qualifications - 5+ years of experience in SRE, production engineering, or infrastructure-heavy backend roles, with strong programming skills in Python, C++, Rust, or Go. - Hands-on experience with cloud infrastructure, production Kubernetes, infrastructure as code such as Terraform, and automated deployment workflows. - Strong Linux and networking fundamentals, including TCP/IP, DNS, TLS, HTTP, and load balancing. - Experience building observability with tools such as Datadog, Prometheus, or Grafana, and using SLOs to guide reliability improvements. - Proven experience diagnosing distributed-system failures, leading production incidents, and improving system performance or resource efficiency. Nice to have - Experience operating web crawlers, search infrastructure, or browser fleets at scale. - Experience with distributed queues, backpressure, workload isolation, or multi-region recovery.