Data Scientist
Current- Development and analysis of deep learning models for Automatic Speech Recognition (Acoustic and Language Model) using Kaldi, DeepSpeech, wav2vec2, Whisper... for different conditions: accents in the language, audio signal with high compression, noisy environments ...- Development of speaker diarization models: spectral, agglomerative clustering, AHC, UIS-RNN... based on speaker embeddings.- Development of audio signal models: voice activity detection, sound event detection, audio KPIs...- Model optimization and deployment into production keeping scale, speed and accuracy in mind.- Pipeline tools: ArgoFlow