To build the best speech recognition models for Siri, Apple needs to use the latest technology in distributed training and the best available data, efficiently training on petabytes of audio data and developing new models to extract useful signals from unprecedented volumes of data.
Requirements
- Experience processing large, complex, unstructured data
- Deep knowledge of distributed data processing frameworks (Beam, Spark, Dask, Ray)
- Evaluation of multiple open source models
- Machine Learning experience a plus
- Speech understanding or generation experience a plus
- Strong data engineering background in speech and/or language/text/dialogue processing field
- Strong software engineering abilities, ideally Python
Responsibilities
- Work with open source tools like PySpark, Jax, Ray and others
- Run data based ablation studies
- Use open source models to extract signals from large volumes of speech data to drive modeling improvements
Other
- M.S. or Ph.D. degree in Computer Science, or equivalent experience
- Real passion for building research demo data solution prototypes and turning them into production quality design/implementation
- Strong interpersonal skills to work well with engineering teams
- Excellent problem solving and critical thinking
- Ability to work in a fast-paced environment with rapidly changing priorities
