Principal Architect - Speech Recognition Job in Reverie Language Technologies Limited

Development of ASR engine using frameworks like DeepSpeech, Kaldi, wav2letter, Pytorch-Kaldi, CMU Sphinx

Assist to define technology required for Speech to Text services besides core engine and to design integration of these technologies.

Work on improvement of model's accuracy and guide the team with best practices.

Lead a team of 3-5 members

Good understanding of machine learning(ML) tools.

Should be well versed in classical speech processing methodologies like hidden Markov models (HMMs), Gaussian mixture models (GMMs), Artificial neural networks (ANNs), Language modeling, etc.

Hands-on experience of current deep learning (DL) techniques like convolutional neural networks (CNNs), recurrent neural networks (RNNs), long-term short-term memory (LSTM), connectionist temporal classification (CTC), etc used for speech processing is essential.

The candidate should have hands-on experience with open source tools such as Kaldi, Pytorch-Kaldi.

Familiarity with any of the end-to-end ASR tools such as ESPNET or EESEN or Deep Speech Pytorch is desirable.

The candidate should have a good understanding of WFSTs as implemented in OpenFST and Kaldi and should be able to modify the WFST decoder as per the application requirements.

Experience in techniques used for resolving issues related to accuracy, noise, confidence scoring etc.

Ability to implement recipes using scripting languages like bash and perl

Ability to develop applications using python, c++, Java