Google is on the lookout for a senior software engineer to "revolutionize" its networking switch infrastructure.

A job listing sees the hyperscaler in need of a Python or C++ proficient engineer with three years or more of experience “with developing large-scale infrastructure, distributed systems or networks, or experience with compute technologies, storage, or hardware architecture.”

The chosen engineer will work on the performance, reliability, and efficiency of the backbone that underpins all of Google's services, with the note stating: “In this role, you will revolutionize how Google builds and operates its data center networks. You will develop the software stack that powers the switches at the heart of Google's global network. By leveraging and extending the open-source technology, you will provide a highly programmable and robust platform for Google's earth-scale software-defined infrastructure.”

Google claimed the U.S. base salary range for the full-time position is $174,000-$252,000, plus bonus, equity, and benefits.

Google's networking architecture is defined by Jupiter, a network fabric encompassing more than 100,000 servers, which the hyperscaler claimed scales to 13 Pb/s of bisectional bandwidth.

Developed in 2015, Jupiter works alongside several software-driven network elements. These include WAN B4, which interconnects global data centers with high-throughput traffic engineering; and WAN B2, which links these data centers to the internet backbone with strict service level agreements (SLAs) for reliable user-facing traffic.

There is also Espresso, which is an SDN controller that operates at the peering edge, dynamically optimizing routes based on real-time latency and availability; and Orion serving as the centralized SDN control plane, managing provisioning and traffic flow across the entire network.

The final piece is Andromeda, which provides network virtualization within the Google Cloud Platform (GCP), creating and managing virtual networks over the underlying physical infrastructure.

Most recently, Google retooled its network to meet AI-era challenges such as extreme scale, stricter reliability expectations, diverse service requirements, and the need for minimal human operation.

To address exponential scalability, it shifted its network from vertically scaling infrastructure to a multishard architecture, where independent network “shards” or independent pieces can scale horizontally, enabling rapid WAN expansion. For constant reliability, it combines shard-level isolation, regional containment, and fast rerouting, reducing outage impact and cutting cumulative outage minutes by reportedly up to 93%.

To meet intent-driven programmability, Google also built a fully software-defined, API-driven network using models, allowing fine-grained, application-aware control over routing, compliance, and performance. Finally, for autonomous purposes, it leverages machine learning – namely graph neural networks – to create a real-time digital twin of the network.