Hey, Satya!

Automatic Speech Recognition Inference

Introduction

Hey, Satya Nadella (and Microsoft), my name is Brice, and I really want to work for Microsoft! I've applied to many jobs, but have had a hard time getting interviews, so I made this website to showcase my capabilities and capture your attention. I'm an an addictions counselor who pursued data science after realizing the mental health service system was not receptive to outcomes-based approaches. I believe that I can bring a unique perspective to Microsoft's vision of empowering the public through data-driven technology for social good.

This website runs a live demo of a scalable automatic speech recognition platform built with a neural network for speech inference that was inspired by my mental health career and a desire to build a system for therapeutic intervention outcomes measurement. My model's cosine similarity score is 72% and word error rate of 21%. Use the inference engine below to test out the model on your own recorded audio files. Then, if you found this impressive, email me at brice@bricewalker.com so we can set up an interview.

Inference Engine

Use the input box to upload a .wav file to run inference on.

Note: Audio files must be in WAVE/WAV format (file extension must be '.wav') and no more than 10 seconds in length. The audio file must be encoded in PCM single channel (mono).The bit rate must be at least 64k, and the sample rate must be 16000Hz (16Khz). The max frequency cannot be more than 16000Hz. You can record using your microphone with a tool like Audacity. Alternatively, you can download the sample files provided to test the model.

Sample files:

Sample File 1:

Sample File 2:

Sample File 3:

Select file for inference:

Thanks for checking out my project. You can demo ASR with Azure Cognitive Services on the ASR tab above and you can click on the visualization or performance tabs to get a more in depth view of the data and model. The about tab also has an outline of my development process. I've included links below to view the code on github or to check out the interactive code notebook on JupyterHub. If you found this impressive, email me at brice@bricewalker.com so we can set up an interview.