AI/ML Speech Algorithm Engineer
Summary
Join Nuance Audio's Audio AI team in Tel Aviv to develop speech AI algorithms (enhancement, feedback cancellation, beamforming, scene recognition) for hearables, carrying them from research and POC through productization on resource-constrained embedded DSP/NPU targets. Core stack is Python with PyTorch/TensorFlow, plus model optimization and real-time embedded deployment.
We are seeking an experienced AI/ML Speech Algorithm Engineer to join our Audio AI team working on cutting-edge speech AI technologies for hearables. You will contribute across the full research-to-productization pipeline — from exploring novel architectures and proof-of-concept work, through productization and deployment on resource-constrained wearable devices. The work moves between two modes: deep research and POC when we're exploring what's possible, and disciplined productization when we're taking a proven approach to a shipping device. You will collaborate closely with AI research scientists, classical DSP algorithm engineers, embedded software engineers, and product teams to translate research outcomes into shipping product features.
What you'll do
• Contribute to speech AI algorithms end to end — speech enhancement, acoustic feedback cancellation, beamforming, acoustic scene recognition — from architecture exploration through to a verified feature on a shipping device.
• Design, train and evaluate deep learning architectures against hard real-time budgets: [target algorithmic latency, MIPS, low footprint on our embedded DSP/NPU]. Quantization, pruning, and other model optimization techniques for deployment are part of the design.
• Build out our evaluation stack: objective and perceptual metrics, listening tests together with our audiology colleagues, and regression testing that tells us whether a model actually improved.
• Generate and curate the data the models need — noise corpora, room impulse responses and acoustic simulation, multi-microphone and binaural recordings.
• Contribute reusable training and testing infrastructure across tasks.
Requirements
• MSc or PhD in Electrical Engineering, Computer Science or a related STEM field, with real depth in deep learning and signal processing. A BSc with equivalent demonstrated depth works too.
• 5+ years building audio or speech ML algorithms, including at least one algorithm you personally carried from research into a shipped product.
• Hands-on model development and production experience in Python, with PyTorch or TensorFlow.
• A track record of deploying models to real-time or embedded targets under latency, memory and compute constraints, and of iterating on the basis of measurements rather than intuition.
Bonus points for us
• Hearing-aid-class signal processing, or experience with hearables and wearable audio products.
• End-to-end audio system understanding: transducers, microphones, audio ICs, acoustics, embedded DSPs and NPUs.
• Familiarity with model optimization techniques — quantization-aware training, pruning, knowledge distillation, and other methods for compressing models without sacrificing performance.
• Cloud-based training environments (GCP or equivalent) and experience with GPU cluster workflows.
• Exposure to regulated or medical-device product development.
• Publications or patents in speech, audio or applied deep learning.
