ML Research Engineer for Video Translation in the Browser

We are seeking a strong ML Research Engineer ready to build from scratch and innovate. You will work on a critically important direction of the speech synthesis service — the video translation technology in Yandex Browser. Our goal is to bring the voice of translated video to a level where the original delivery, naturalness and emotionality, intonation patterns, and even singing style are preserved.

Your developments will provide millions of users with the most natural and emotionally accurate translation of video content, breaking down language barriers and making content even more accessible and vibrant. You will work on the most challenging areas of ML, solving tasks related to transferring and preserving all the nuances of speech.

Tasks you will face

Development and Implementation of ML Models For accurate analysis of the source audio track and transferring intonation, emotional coloring, and speech pace into the synthesized translation.

Work on Reproducing Non-Verbal Communication Elements Laughter, sighs, pauses, interjections, and other natural speech features are critically important for maximum translation naturalness.

Research and Creation of Algorithms For successful copying and synthesis of singing in the target language.

Analysis and Improvement of Algorithms To minimize pronunciation errors in speech synthesis.

More about ML at Yandex — in the channel Yandex for ML

We expect you to

• Confidently master Python, have excellent knowledge of machine learning frameworks, specifically PyTorch
• Have practical experience with distributed training and large models
• Have broad technical expertise in NLP and are ready to implement new technologies from scratch
• Want to dive into the field of speech synthesis, ready to understand both the theory and the engineering details of implementation
• Are capable of keeping up with the development of machine learning and turning research ideas into reliable, working code

Will be a plus

• Have worked with multimodal or generative models
• Have experience in the field of TTS/VC (Text-to-Speech/Voice Conversion).