Software Engineer, Models

Meter•San Francisco, CA

3d•$220,000 - $278,000

About The Position

The Problem We're Solving The internet runs the world. But the networks it runs on haven't meaningfully changed in decades. They're brittle, complex, and surprisingly hard to manage at enterprise scale. Meter is building the first truly autonomous networking platform: a physical AI system that learns from real network behavior and resolves issues before they ever surface as problems. What makes this possible is data. Because we own the entire enterprise networking stack — custom hardware, firmware, orchestration software, and API — we have access to real-world telemetry that no one else can see. But raw telemetry isn't enough. The real signal lives in the minds of expert network engineers: the way they read a dashboard, form a hypothesis, and trace a problem to its root cause. Our ambition is to embed that judgment into our models at scale and use it to transform one of the largest job categories in tech. What This Role Is Nobody is building what we're building, and a big reason for that is data. If we can systematically capture how great network engineers think and encode it into our models, we build something genuinely unreplicable. That's what this role is about. You'll own the systems and tooling that turn Network Engineer expertise into high-quality training data for our foundation models. This is a 0-1 role: you'll spend real time alongside Network Engineers learning how they think before you write a line of code, then build fast, iterate faster, and scale what works. You'll know you've succeeded when the Network Engineering team can generate training data independently, at volume, and model performance moves as a result. Why Meter, Why Now Ambitious companies and enduring institutions — Bridgewater, Lyft, Reddit — rely on Meter to keep their employees and locations online and productive. We had to build everything from the ground up to get here: our own enterprise hardware, intuitive software, and streamlined operations. The models team is where the next chapter gets written. We have the stack, the data, and the customer base. What we're building now is the intelligence layer that makes enterprise networking autonomous. This is the kind of problem that comes around once — and we're at the beginning of it.

Requirements

Move fast and know when "good enough to learn from" beats "perfect" — and when it doesn't
Can prototype quickly and have the skill and fortitude to bring it to production when the time is right
Build for the end user: you care whether the people using your tools actually enjoy using them
Are curious about domains outside engineering — you'll need to learn how network engineers think, not just what they do
Communicate clearly with both engineers and domain experts
Thrive with high autonomy and are energized by defining the work, not just executing it

Nice To Haves

Experience with data labeling, active learning, or annotation frameworks
Understanding of real‑time telemetry and large‑scale time‑series data
Background building internal tools or developer-facing products

Responsibilities

Build and Own Labeling Tooling
Design, implement, and iterate on the tools that network engineers use to annotate issues, log diagnostic reasoning, and curate data for model training and evaluation
Build interfaces that feel natural to domain experts: low friction, high signal, and easy to use without engineering support
Partner closely with network engineers to continuously learn what's working and what isn't, and update the tooling accordingly
Move Fast and Iterate
Ship working prototypes quickly, treat v1 as a learning artifact, and bring proven approaches to production
Own the full cycle: design, build, deploy, observe, revise
Be comfortable with ambiguity. You're defining the playbook, not following one
Scale the Data Operation
Set up metrics and feedback loops that measure tooling effectiveness and its downstream impact on model quality
Monitor and troubleshoot data pipelines and tooling in production
Help define best practices for labeling workflows and data governance