About the job
About Meter
At Meter, we believe that networking is at the heart of technological advancement. We have innovatively unified the entire networking stack and are now on a mission to make it autonomous.
Our team is developing a cutting-edge neural network-driven system designed to analyze raw computer networks, enabling us to address all networking challenges. As outlined on Meter.ai, we are creating models within a closed-loop system that utilizes real-time telemetry, logs, and network events to autonomously troubleshoot issues, enhance performance, and resolve challenges.
To achieve this, we require not only exceptional models but also robust infrastructure that ensures our models have clean, versioned, and low-latency access to the necessary data throughout training, evaluation, and deployment phases.
Why this Role is Essential
Each Meter network deployed in the field serves as a valuable data source for our Models team. However, without meticulous infrastructure design, this data risks becoming fragmented, outdated, or inconsistent. In this role, you will ensure that such pitfalls are avoided. You will be responsible for the core data interface that drives our model development, experimentation, evaluation, and real-time inference.
This position is fundamental and offers a significant impact. Your contributions will shape the speed at which we can train new models, the reliability of their evaluations, and their seamless operation across hundreds of real-world networks. You will collaborate closely with modelers to deliver systems that are elegant, scalable, and robust.
Your Responsibilities
Design and implement the Models API: a unified interface for accessing training, evaluation, and deployment data across raw, transformed, and feature-engineered layers.
Ensure backward compatibility and feature versioning across continually evolving schemas.
Develop scalable pipelines to ingest, transform, and serve petabytes of data across Kafka, Postgres, and Clickhouse.
Create CI/CD workflows that evolve the API in tandem with changes to the underlying data schema.
Facilitate fine-grained querying of historical and real-time data for any network, at any point in time.
Help establish and promote the principle of 'smart data, dumb functions': maximizing operations in the data layer to minimize downstream code complexity.
Collaborate with modelers to co-design training frameworks that optimize performance.
