Research · Santa Clara, California (on-site) · Full-time
Machine learning engineer, neural audio
Own the learned models at the center of our audio chain — suppression, separation, prediction — and make them run on a vehicle NPU inside a frame-rate budget.
title: Machine learning engineer, neural audio location: Santa Clara, California (on-site) team: Research employment: Full-time posted: 2026-08-03 summary: Own the learned models at the center of our audio chain — suppression, separation, prediction — and make them run on a vehicle NPU inside a frame-rate budget.
About the role
You will own learned models at the center of our in-cabin audio chain: suppression, separation, and prediction for echo cancellation and road-noise cancellation. The models you train do not end in a paper — they end on an automotive NPU, holding a frame-rate promise, in a cabin that never sounds the same twice.
This is a role at the intersection of applied machine learning and real-time audio. The strongest candidates have lived on both sides.
What you would work on
- Learned components of a hybrid audio chain: the neural core between deterministic DSP stages
- Data pipelines built from cabin recordings — coverage design, augmentation, and the honesty of the test set
- On-device constraints as first-class design inputs: streaming inference, quantization, and memory budgets
- Evaluation against strong classical baselines, with a trained ear in a real seat as the final metric
What we expect
- Applied experience in speech enhancement, acoustic echo cancellation, source separation, or active noise control
- Fluency in PyTorch and comfort reading and writing the C or C++ that deployment eventually demands
- Respect for classical signal processing — you should know what an adaptive filter does well before replacing one
- A low tolerance for unverified claims in your own work, and in others’
What we do not require
A PhD. Publications. A specific institutional background.
Logistics
- On-site in Santa Clara, California. We do the work together, in person.
- Full-time, full benefits
- Compensation is competitive. Exact bands on request.