Responsibility: Research and development of state-of-the-art speech technologies for low-resource languages using semi-supervised methods. Development of ASR systems using frameworks like Kaldi, ESPNET, DeepSpeech, Athena or FairSeq using PyTorch or Tensorflow. Deployment and maintenance of ASR core engine for multiple domains. Improvement of model performance. Guiding the team with best practices. Good to Have: Experience in AI/ML/DL with a good understanding of machine learning tools. Ability to train, test and diagnose models. Experience in techniques used for resolving issues related to accuracy, noise, confidence scoring etc. Experience in designing and developing scalable deep learning systems. Experience in building speech technologies such as ASR, TTS, speaker verification, speech enhancement, far field speech recognition, speaker diarization, etc. Should be well versed in classical methodologies like hidden Markov models (HMMs), Gaussian mixture models (GMMs). Familiarity with weighted finite state transducers, decoding algorithms, etc. Hands-on experience with current deep learning concepts such as RNNs, LSTM, GRU, CTC, etc. Familiarity with latest architectures like encoder-decoder models, attention mechanism, transformer models etc. Hands-on experience is a plus. Skills: Hands-on PyTorch or Tensorflow experience is desirable. Ability to implement recipes using scripting languages like bash. Ability to develop applications using python, C++, Java Good to Have: Peer-reviewed publications in conferences/journals with a good impact factor.
-
Seniority Level:
Engineering & Technical
-
Industry:
Internet and IT
-
Employment Type:
Full Time