Machine Learning Engineer — Multi-Sensor Time-Series & Signal Processing

About the role We're building an end-to-end machine learning pipeline that turns raw, multi-sensor wearable data into real-time event detection and quantitative estimates. We're looking for an experienced ML/DSP engineer who is comfortable owning the full path from raw sensor streams to a trained, validated model. You'll work with synchronized multi-modal time-series data (strain/pressure sensors, audio, and motion/IMU), build deep learning models for event classification and regression, and help us bridge the gap between synthetic/lab data and real-world signals. What you'll do Design and train deep learning models for time-series (CNN + LSTM / sequence models) with multi-task heads (classification + regression) Build and harden the data pipeline: ingestion, calibration, resampling, multi-rate time alignment, and signal-quality checks Engineer features across modalities (spectral/audio features, filtering, windowing) Develop synthetic data generators and augmentation strategies Run rigorous evaluation: cross-validation, ablations, and held-out real-world validation Aggregate per-window predictions into reliable session-level metrics Required skills Strong Python and the scientific stack (NumPy, SciPy, pandas) PyTorch (or TensorFlow) for sequence modeling — CNNs, LSTMs/RNNs, multi-task learning Digital signal processing: filtering, resampling, STFT/spectrograms, log-mel features, windowing Experience with multi-sensor / time-series data: synchronization across different sample rates, sensor calibration, signal-quality assessment Solid ML fundamentals: handling class imbalance, loss weighting, scheduling, early stopping, cross-validation, and ablation studies Clean, modular, config-driven code with tests and reproducibility (Git, packaging, seeding) Nice to have Experience with wearable / biomedical / embedded sensor data Familiarity with the synthetic-to-real domain gap and transfer/validation strategies Audio ML or acoustic event detection background Comfort collaborating with hardware teams on data collection and ground-truth labeling To apply Please include: - Examples of past time-series or sensor-based ML work (repos, papers, or case studies) - Your experience level with PyTorch and DSP specifically - Any history of studying at universities in native-English-speaking countries (e.g., US or UK), as strong written and spoken English is important for close collaboration Please begin your proposal with details of any studies you completed at universities in native English-speaking countries (e.g., the US or UK). Proposals that don't start with this will be skipped.