Machine learning for speech, audio and hearing. Marie Skłodowska-Curie Fellow at the Deep Hearing Lab (University of Zurich / University Hospital Zurich); PhD in Statistical Science and Signal Processing, UCL.
I build models that connect acoustic signals, auditory-nerve representations and speech perception — and methods that compare their structure instead of collapsing it into a single score.
Selected work
- SimsACues_ASR — neurogram simulation, two-stage phoneme recogniser and Gromov–Wasserstein analysis for auditory nerve disorders (installable Python package, tested, CI)
- CI_Outcome_Pred — predicting speech outcomes after cochlear implantation from pre-operative data on a national registry (TabPFN, conformal prediction intervals, temporal validation)
- EMD-Stochastic-Embedding-for-PD-Speech — Parkinson's disease detection from speech via Gaussian-process embeddings of EMD
- GapDetection — open-source gap-in-noise detection test, in use with patients
- EMD-MFCC-SVM-Speech — EMD-based features for robust speaker verification against synthetic speech
- SpiN — hearing-loss risk profiles with state-space models
Tools · Python · PyTorch · TensorFlow · R · MATLAB · optimal transport · signal processing