> Markdown version of [/jobs/ext/4790786-systems-engineer-maps](https://www.wearedevelopers.com/jobs/ext/4790786-systems-engineer-maps). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Systems Engineer, MAPS - **Company:** CloudFlare - **Location:** New York, NY, United States - **Contract:** Permanent contract - **Skills:** Artificial Intelligence, Systems Engineering, Unix, Profiling, Computer Networks, Databases, Software Debugging, Linux, Domain Name System (DNS), Protocol Buffers, Linux System Administration, Network Control, Data Streaming, System Programming, Load Balancing, Concurrency, Perf (Linux), Data Analytics, Cloudflare, Vertica, Data Pipelines - **Published:** November 19, 2026 - **Apply:** https://www.jofdav.com/jobs/60601857-systems-engineer-maps ## About the Role Strong systems programming experience in Rust, Go, or C, with a solid grasp of memory layout, concurrency, and performance-critical code paths. Experience designing and operating latency- and resource-sensitive server infrastructure or high-throughput data pipelines at scale. Strong fundamentals in Linux systems internalsprocesses, cgroups, sockets, and how the kernel accounts for compute, memory, and I/O. Proven ability to work across organizational boundaries: reading unfamiliar codebases, aligning with other infrastructure teams, and driving complex integrations to completion. A pragmatic approach to owning outcomesfrom kernel-level data collection and wire protocols (protobuf/Cap'n Proto) all the way to the analytical tables consumers query., Hands-on experience writing and verifying eBPF programs, or deep familiarity with the Linux networking stack and scheduler. Experience designing schemas and query patterns for distributed columnar databases (particularly ClickHouse) at high ingestion rates. Background in profiling, benchmarking, and debugging performance bottlenecks in production Linux environments (perf, PMU counters, flamegraphs). ## Description At Cloudflare, were not looking for people who wait for a polished roadmap; were looking for the builders who see the cracks in the Internet that everyone else has simply learned to live with. We value candidates who have the instinct to spot a "normalized" problem and the AI-native curiosity to create a solution using the latest tools. Our culture is built on iteration, leveraging AI to ship faster today to make it better tomorrow, while ensuring that every improvement, no matter how small, is shared across the team to lift everyone up. If youre the type of person who values curiosity over bureaucracy, and that AI is a partner in solving tough problems to keep the Internet moving forward, youll fit right in., Cloudflares architecture is built on a homogeneous edge: every server in over 330 cities runs every servicefrom L4/L7 proxies (FL2, Pingora) and DNS to the Workers runtime, cache, and storage. While this design maximizes hardware utilization, it creates a difficult distributed attribution problem. Machine-level telemetry tells us how loaded a server is, but not which product, customer, or upstream service is responsible for that load. MAPS (Metrics and Analytics for Product Segmentation) is a newly formed team building the core instrumentation, collection, and data platform to solve this from the ground up. We are architecting a system to continuously measure CPU, network bandwidth, memory, and storage utilization across the entire global fleet without samplingattributing resource consumption both to the underlying system service and to the specific product, account, and zone being served. The data MAPS produces serves two critical functions at Cloudflare: Real-Time Control Plane: MAPS is the primary telemetry input for Cloudflares next-generation global load balancer, shifting traffic management from open-loop heuristics to closed-loop routing driven by real-time, per-service resource capacity and consumption (within seconds). Fleet Economics & Capacity Planning: MAPS provides the long-term analytical dataset in ClickHouse used to map product demand forecasts to physical hardware requirements, calculate true cost-to-serve per product, and drive multi-year infrastructure investments. What Youll Do You will work across the full lifecycle of a greenfield, high-throughput telemetry system running on every machine at the Cloudflare edge: Host-Level Instrumentation ("Outside View"): Build and maintain low-overhead Linux daemons (netacctd, maps_collector) in Rust and C that use eBPF (cgroup_skb) and cgroup v2 accounting to measure per-service resource consumption directly from the kernel. High-Performance Client Libraries ("Inside View"): Architect non-blocking, zero-allocation reporting SDKs in Rust and Go used by Cloudflares highest-throughput data-plane services to report per-request resource usage and product attribution over local Unix domain sockets. Edge Aggregation & Streaming: Build the per-machine collector daemon that ingests millions of events per second, performs bounded in-memory time-window aggregation, serves low-latency gRPC streams to the global load balancer, and flushes batch rollups to ClickHouse. Reconciliation & Correctness: Solve complex attribution problemssuch as reconciling kernel-level cgroup measurements against application-reported metrics, and attributing resource usage when one multi-tenant service executes work on behalf of another. Cross-Team Architecture: Partner directly with the engineering teams behind Cloudflares core proxies, runtimes, and storage engines to instrument their critical paths without degrading tail latency.