Research Scientist, On-Device Speech Recognition
Facebook Reality Labs (FRL) is the world leader in the design of virtual and augmented reality systems. Come work alongside expert engineers and research scientists to create the technology that makes VR and AR pervasive and universal. Join the adventure of a lifetime as we make science fiction real and change the world.
We are seeking a Research Scientist to support the development of state-of-the-art technology that enables on-device speech recognition for AR/VR systems. The successful candidate will be part of our efforts to architect, design and implement the software and hardware platforms for this activity and will be part of a team that includes algorithm, user experience, software, firmware and ASIC experts. The ideal candidate will have research experience on the full stack of speech recognition pipeline including audio, speech, and language modeling and understand the technical challenges of porting it on-device.
This is a full-time position based in either our Menlo Park, CA or Redmond, WA offices.
- Develop and optimize machine learning models for on-device speech use-cases, including speech recognition, natural language understanding, and speech synthesis. Such models will surpass state-of-the-art metrics for accuracy, computation complexity, model sizes and power consumption
- Work with architecture research teams to develop software/hardware acceleration solutions
- Work with algorithm research teams to map the speech recognition pipeline to hardware implementations, create cost-benefit analysis and estimate power and performance
- Build integrated speech recognition systems and perform evaluations
- 2+ years of experience and Ph.D. in Computer Science, Electrical Engineering or equivalent field
- Experience in machine learning, deep learning, and/or speech recognition
- Software design and programming experience in C/C++ for development, debugging, testing and performance analysis
- Experience writing research code with the standard python stack (NumPy, SciPy, etc.), speech recognition toolkits (e.g., Kaldi) and at least one modern deep learning library (e.g., PyTorch, TensorFlow)
- Experience crossing multi-disciplinary boundaries to drive optimal system solutions
Back to top