Audio AI Research Scientist — Audio Source Separation

Bose Corporation

2d•$165,400 - $227,500

About The Position

Bose is hiring a Research Scientist to develop novel AI-powered audio processing algorithms that run in real time on physical devices. The role focuses on algorithms that isolate, extract, or suppress individual sound sources from microphone recordings, potentially using additional sensor signals. Applications include voice pickup, hearing augmentation, dynamic awareness, agentic perception, and other audio experiences. The scientist will work with experts in machine learning, digital signal processing, software engineering, and psychoacoustics to prototype and implement new algorithms. The work includes addressing practical deployment constraints such as perceptual quality, latency, memory, and computational requirements, and collaborating with DSP and embedded software teams to bring models to real-time demonstrations and product platforms.

Requirements

Graduate degree (MS or PhD) in machine learning, electrical engineering, computer science, or a related field, or equivalent research experience.
Deep hands-on experience in audio source separation, target sound extraction, speech enhancement, or related neural audio processing.
Comfort with complex, open-ended problems that require experimentation and iteration.
Strong knowledge of audio DSP, time-frequency methods, and microphone array or spatial signal processing.
Experience training and evaluating audio models in PyTorch.
Understanding of objective and perceptual evaluation methods and the tradeoffs involved in real-time, low-power inference.
Strong communication skills.

Nice To Haves

Proficiency in Python.
Familiarity with C/C++.
Experience with multimodal or sensor-conditioned separation, ONNX/TFLite, or embedded deployment.
Publications at conferences such as ICASSP, Interspeech, NeurIPS, or similar.

Responsibilities

Research, prototype, train, and evaluate novel source separation and extraction algorithms for speech and other real-world sounds.
Develop approaches for single- and multi-microphone recordings, including spatial cues and optional auxiliary sensors.
Investigate robust separation in challenging conditions, including overlapping talkers, noise, reverberation, and motion.
Evaluate perceptual quality, separation performance, generalization, latency, memory, and computational requirements.
Collaborate with DSP and embedded software teams to deploy models in real-time demonstrations and product platforms.
Contribute to patents and conference papers.