ML Compression Engineer

The Compression Company•San Francisco, CA

About The Position

You will build the compression technology at the core of The Compression Company. You'll design and train neural compression models, improve compression ratio while preserving quality and downstream utility, and make our training system fast and repeatable so we can produce SOTA codecs for new datasets with minimal manual effort. Your work will directly drive product performance and the pace at which we can ship into new customer environments.

Requirements

Trained and shipped ML models end-to-end
Strong in PyTorch
Comfortable owning training + evaluation
Enjoy turning research work into usable systems

Responsibilities

Design and train neural compression models
Improve compression ratio while preserving quality and downstream utility
Make our training system fast and repeatable to produce SOTA codecs for new datasets with minimal manual effort
Improve rate/quality performance for real customer datasets
Build automated training runs that explore model + training settings and select strong candidates
Extend our evaluation suite to reflect real user needs, not just offline metrics
Improve reproducibility and speed across the training + benchmarking loop