C/C++ is the language, named in most job posts and in nearly all outside big tech. The C side is for the kernel, drivers and anything that must be small and predictable; the C++ side is for the large engines, storage systems, databases and network software where the code base runs to millions of lines. Rust is beginning to appear, in a small share of job posts, most at the security companies whose packet-processing code must be both fast and safe, and in big tech. Python and the shell are assumed for the tools and tests around the code, and Assembler appears at the processor makers, where the engineer reads what the compiler produced. ARM and x86 name the processor families the code runs on.
The tools of the trade appear by name because they are the skill. GDB is the debugger, and knowing how to read a crash in it is the mark of a systems engineer. strace and tcpdump show what a process and the network are doing, Wireshark inspects captured traffic, WinDbg is the Windows debugger at Microsoft, and QEMU and KVM run a virtual machine to test kernel code without a second computer.
The Linux Kernel is the second most-named skill in the whole role. It is the core of the operating system, and systems engineers work inside it: fixing it, tuning it, adding to it, and keeping a company's version of it current. Linux Device Drivers are the pieces of the kernel that talk to a particular piece of hardware, a network card, a storage controller, a GPU, and writing them is the most common single task in the kernel job posts. Linux BSP and Embedded Linux are the board support package and the slimmed-down Linux that make the kernel run on a particular machine, a server, a switch or a storage array, and they are named together on a large share of job posts. I/O Device Drivers, DMA, PCIe and SoC are the pieces of the hardware a driver writer deals with. eBPF is the newer way of running safe code inside the kernel without changing it, and it appears at the storage and server makers and the security companies. Secure Boot appears where the machine must prove its software has not been tampered with.
This work is heaviest at the server and storage makers, where it is most of what a systems engineer does, and at the operating-system and virtualisation companies, which build Linux itself. The processor makers ask for it too, for the drivers, Compilers and GPU libraries that make their own chips usable, with NVIDIA CUDA named at the GPU end.
The server and storage companies build the machines and software that hold the world's data, and their job posts name the storage stack. File Systems is named in a fair share of kernel job posts and in a large share at the storage makers: the code that turns a disk into files and folders and keeps them safe when the power fails. NFS and SMB are the protocols for serving files over a network, NVMe, SCSI and Fibre Channel the ways a machine talks to its disks, RAID the way disks protect each other, and NAS and SAN the two shapes of a storage system. Ceph is the open-source distributed store. Database engines are a cousin of this work: query execution, storage engines and distributed transactions in C++, at the database companies, which are among the more junior-friendly employers in the role.
The cloud is built on virtual machines and containers, and systems engineers build that layer. KVM is the Linux hypervisor, Microsoft Hyper-V the Windows one, QEMU the machine emulator they use, and OpenStack the open-source cloud built on them. These appear at the operating-system companies, the server makers and the processor makers. Cloud and containers are tagged on a large share of job posts, but the meaning differs from other roles: Kubernetes, Docker, AWS, Azure and GCP are named because the systems engineer builds the storage, networking and virtualisation that the cloud runs on, or packages a storage or network product to run inside the public clouds, as the storage and security companies now do. The tag is highest at the server and storage makers and the security companies.
The network and telecom job posts are a world of protocols, and they are named one by one. IP and TCP/IP are the foundation, with UDP, IPv6, DNS, DHCP, HTTP/HTTPS, NAT and SNMP beside them. Layer 2 and Layer 3 are the switching and routing halves of a network operating system, and Ethernet, VLAN and VXLAN are the layer 2 technologies. BGP, OSPF, MPLS, Multicast and IGMP are the routing protocols that decide where packets go, and QoS decides which packets go first. These are the whole of the job at the enterprise networking equipment makers, which build the switches and routers, and at the telecom equipment makers, where 5G and LTE name the mobile stacks. The engineering services firms that serve these companies ask for the same protocols, with the most GDB and the most testing.
The security companies whose products sit in the path of customers' traffic, firewalls and cloud security gateways, need code that inspects every packet without slowing the network, and their job posts are the most demanding form of network work. DPDK is the toolkit for processing packets in user space at the speed the hardware allows, bypassing the kernel, and it is named most at these companies and the server makers. IPSec, VPN, Firewalls and DNS are the things being built, and Rust appears here more than anywhere else. Security is tagged on most job posts at these companies, and it means encryption, secure boot and the hardening of the data plane itself.
Big tech is the largest single home for the role, and its job posts read differently. Amazon's are large-scale services and distributed systems in C++, at the mid level, with the systems vocabulary of kernel and drivers largely absent because the work is above the kernel. Microsoft's are Windows, Azure infrastructure and storage, with WinDbg and Microsoft Hyper-V. Google's are infrastructure, storage and networking for its own data centres, and it is the most junior-friendly systems employer in big tech. Build and release is tagged on most big tech job posts, because the engineer owns the whole route from code to a running fleet.
Systems code is built with its own toolchain and tested on real or virtual hardware, and build and release is the most common extra tag in the whole role, on a large share of job posts. It means the build systems that compile a kernel or a network operating system, Jenkins, Azure DevOps, GitHub Actions and GitLab CI/CD where a pipeline carries the result, and Git throughout. Testing is tagged on a fair share, most at the engineering services firms and the server makers, and it means test automation in Python with pytest, hardware in the loop, and the lab equipment that exercises a switch or a storage array. Ansible and Terraform appear in a few job posts for setting up test fleets.
A small share of job posts reach into neighbouring work. Machine learning on the device, tagged most at the processor makers, means making models run fast on the company's own silicon, with NVIDIA CUDA. Wireless protocols appear at the telecom and home-router companies. ASIC, SoC and chip-design tooling appear where the systems engineer sits beside the hardware team. In the ordinary systems job these are a plus.
Underneath all of it sits the way systems software is built: Git and code review, long-lived branches for each hardware release, careful performance measurement before and after every change, and debugging that can take days. Job posts count these as given and rarely list them as skills.
A systems engineer who writes C/C++ well, knows the Linux Kernel and can write a Linux Device Driver, is at home in GDB, understands TCP/IP and the network stack, has worked on a File System or a storage protocol or a routing protocol, and can bring up Embedded Linux on a new board, meets the core of nearly every job post. The variations belong to the employer. The server, storage and database makers want the kernel, drivers, File Systems, NFS, SMB and NVMe, packaged increasingly for Kubernetes and the public clouds, and they are the most senior. The networking and telecom equipment makers want Layer 2 and Layer 3, BGP, OSPF, MPLS and QoS on an Embedded Linux base. The security companies want DPDK, IPSec and Rust for packet processing at line speed. The processor makers want drivers, Compilers, CUDA, KVM and Assembler for their own chips, and they are among the more junior-friendly. Big tech wants large-scale C++ services and the infrastructure of its own data centres. Across all of them, the engineer who understands what the hardware is doing underneath the code is the one every job post describes.