The Speak team is Spotifies in-house text-to-speech (TTS) team, supporting products like DJ, AI Voice Translation, as well as the development of exciting new unreleased products. We focus on building world class speech technologies that can power the next generation of personalized generative voice products at scale.

What You’ll Do

• Build large-scale speech and audio data pipelines using frameworks like Google Cloud Platform and Apache Beam
• Work on machine learning projects powering new generative AI experiences and helping to build state-of-the-art text-to-speech models
• Learn and contribute to the teams best practices and techniques for building data pipelines for large scale generative models, including cleaning, filtering, classifying and labelling
• Collaborate with other engineers, researchers, product managers and stakeholders, taking on learning and leadership opportunities that arise
• Deliver scalable, testable, maintainable, and high-quality code
• Share knowledge, promote standard methodologies, making your team the best version of itself through mentorship and constructive accountability.

Who You Are

• You have Data Engineering experience and you know how to work with high-volume, heterogeneous data, preferably with distributed systems such as Hadoop, BigTable, Cassandra, GCP, AWS
• You have experience building clean, high quality datasets for training large scale machine learning models, a focus on audio data is preferred
• You have experience with one or more higher-level Python or Java based data processing frameworks such as Beam, Dataflow, Crunch, Scalding, Storm, Spark etc
• You have strong Python programming abilities. You might have worked with Docker as well as Luigi, Airflow, or similar tools
• You care about quality and you know what it means to ship high quality code
• You have experience managing data retention policies
• You care about agile software processes, data-driven development, reliability, and responsible experimentation
• You understand the value of collaboration and partnership within teams
• You have experience in developing datasets tailored for training high-performance machine learning models.
• Familiarity with generative models or audio-based machine learning applications is highly desirable.
• You are proficient in cleaning, filtering, and evaluating dataset quality, leveraging both pre-trained and in-house machine learning models, as well as human evaluation techniques, to ensure optimal quality.

Were You’ll Be

• We offer you the flexibility to work where you work best! For this role, you can be within the UK region as long as we have a work location.
• This team operates within the GMT time zone for collaboration.