ASR decoder customization

I am looking for a machine learning engineer with extensive knowledge in ASR decoder such as CTC recognition algorithm C++, and python. We have a C++ decoder library/module that we would like to modify. The task itself is not hard, but you must be familiar with machine learning algorithm and C++. This is the link to the decoder. [login to view URL] INSTRUCTIONS the decoder is written in C++ and the entire project is under folder /swig. [login to view URL] is the code we use to compile the decoder to python modules. [login to view URL] is the python interface that we use to call the decoder. All the functions are calling the swig_decoder module. The current decoder returns the whole transcript. We instead, we need to extract word level confidence and word level timestamps. If you read the above contents, when you apply for this project, please address your favorite tea. REQUIREMENTS: modify the decoder code package to support word level time stamp and word level confidence. modify the swig_wrapper to support interface to return word level time stamp and word level confidence as a class