Signal Processing Expert (Audio ML)/ Data Scientist
Engagement Type: Consulting Contract (approximately 8 weeks); full-time preferred, open to part-time
Location: Remote; must have time zone overlap with Eastern US (EST) / Europe (CET)
Compensation: Competitive, commensurate with experience
Start: Jan 5
Hyfe is seeking an experienced Signal Processing Expert/ Data Scientist to design, train, and validate machine learning models for real-time cough detection, as well as to inform important product design decisions. This role will focus on developing efficient, on-device–deployable models trained on a large corpus of annotated, continuous audio data. The engagement is time-sensitive and well-scoped, with a clear deliverable-oriented mandate.
Responsibilities
- Design and implement end-to-end ML pipelines for audio-based event detection (cough focus), including training, validation, and reproducibility
- Develop and validate real-time, on-device models under strict compute, memory, and latency constraints
- Build Python-based training and data pipelines using modern ML frameworks and large, annotated audio datasets
- Collaborate with engineering and product teams to inform system requirements and support deployment with clear technical documentation
Required Qualifications
- Demonstrated expertise in audio signal processing and acoustic feature extraction
- Strong experience building ML models for event detection in continuous time-series or audio data
- Proficiency in Python required
- Hands-on experience with TensorFlow and/or PyTorch
- Proven ability to work effectively with large datasets, including efficient data pipelines and experimentation workflows
- Experience designing models for computationally constrained or embedded/on-device environments
Preferred Qualifications
- Prior work on audio-based biomedical, health, or wearable sensing applications
- Experience with real-time or low-latency inference systems
- Familiarity with model compression, optimization, or edge deployment techniques
- Background in embedded systems or DSP-constrained environments
