Senior Machine Learning Engineer (ASR) (all)

Sonia creates and deploys AI-enhanced solutions that improve doctors’ work, patient care, and healthcare system efficiency. The Senior Machine Learning Engineer (ASR) will develop, evaluate, and deploy speech recognition systems for clinical applications, curate datasets, build audio pipelines, and collaborate on MLOps and cross-functional product development.

Responsibilities:

• ASR Model Development: Fine-tune, evaluate, and deploy state-of-the-art models in speech recognition and audio processing (e.g., Whisper, wav2vec)
• Data Curation & Annotation: Collect and curate custom ASR datasets, including data sourcing, annotation pipeline setup, quality control, and alignment/segmentation procedures
• Audio Pipelines: Build and maintain robust data pipelines and audio preprocessing workflows for clinical environments
• MLOps Collaboration: Work closely with the MLOps team to ensure continuous training, monitoring, and seamless deployment of models in production
• Experimentation: Design and conduct experiments to validate new approaches, datasets, and architectures to improve accuracy in noisy or specialized medical settings
• Cross-Functional Impact: Collaborate with product managers and developers to translate complex speech solutions into production-ready healthcare tools

Requirements:

• Master's degree in Computer Science or Engineering
• Typically 5-8+ years of experience in ML engineering
• Strong programming skills in Python (working with production code and deploying models in production) and ML frameworks (PyTorch, TensorFlow or Jax)
• Direct experience with ASR models (Whisper, wav2vec, …), VAD, alignment and diarisation,  and complex speech/audio processing pipelines
• Extensive experience with transformer-based architectures and deep learning models
• Practical experience with MLOps pipeline components such as Docker, MLflow, W&B, DVC, or Kubernetes
• Knowledge of multilingual or domain-specific modeling (specifically medical, legal, or other specialized terminologies)
• Experience with distributed training systems for large-scale model optimization
• Full proficiency in English and German