ML Engineer in the Speech Synthesis Data Team

We are looking for an experienced data and ML engineer for the speech synthesis data team. The team is engaged in video translation, creates audiobooks, and develops the voice of Alice. In synthesis, an era has begun of transitioning from low resource (even for major languages) to big data and pre-training. New models allow you to sing famous songs in your voice and speak any phrase using just a few seconds of your voice. The foundation of quality for these models is hundreds of thousands of hours of high-quality audio data and corresponding texts, which we need to collect.

What tasks await you

Working with data You will be developing a system for storing truly big data and providing access to it for ML developers. At your disposal will be petabytes of audio, which need to be stored efficiently and processed quickly.

Data mining You will be improving the throughput of current data collection pipelines and scaling them to support multiple languages, working with heterogeneous sources, and developing processes for mining audio data.

Data quality assessment You will be working with processes for assessing data parameters, developing and applying ML models for detecting noise, music, multiple voices, synthetic speech, text-audio mismatches, and language detection. These assessments will allow us to filter data and make our synthesis the best in the world.

Learn more about ML at Yandex — on the channel Yandex for ML

We expect that you

• Write in Python
• Have built data collection pipelines for ML
• Understand how to work with large volumes of data
• Have applied in practice, or better yet, trained ML models
• Are motivated and ready to dive deep into the field