About the Role
Roku is building a world class voice system that is used by millions of Roku users. The Roku Voice team is looking for ambitious experienced machine learning engineers with a background in one or more of the following areas: artificial intelligence, natural language understanding, machine learning, automated speech recognition, conversational systems and building large scale production systems. You have a once in a lifetime opportunity to contribute to building the very core of the Roku Voice product and be part of a world class team.
What you'll be doing
1. Design and develop software and algorithms for Roku’s state of the art voice system
2. Using machine learning expertise, design and develop software components for high availability and high performance cloud solutions
3. End-to-end responsibility from developing a proof of concept to production
4. You have a strong ML programming background with a lot of hands-on experience in building large scale production systems
5. You have a strong algorithmic background and like intellectual challenges and solving complex problems
6. You like working across teams and pulling in the best talent from the organization to achieve your goals
7. You contribute with new ideas and evaluate multiple solutions with your peers before settling down on specific solutions
8. You contribute with new ideas and evaluate multiple solutions with your peers before settling down on specific solutions
9. You are self-driven, willing and able to take complete ownership of initiatives, and make pragmatic technical decisions
We're excited if you have
10. 5+ years hands on experience in building challenging production systems, especially in commercial Speech Recognition systems
11. Strong CS fundamentals, with the ability to write algorithms with ease
12. Deep understanding of DNN/HMM based systems like Kaldi
13. Experience in ASR systems with modern End2End architecture
14. Hands-on experience in Acoustic Modelling with different Neural Network architectures and frameworks like Pytorch and TensorFlow
15. Experience in language model training and adaptation pipeline. Experience in lattice rescoring with RNN or LLM models
16. Deep understanding of decoding algorithms; WFST based decoders is preferred
17. Experience with Transformer based G2P models and training them for different languages
18. Strong background in C++ and Python, and hands-on experience with libraries like Pybind11
19. MS in Computer Science or related field; a Ph.D. in CS or related fields is preferred
#LI-AM3
Benefits
Roku is committed to offering a diverse range of benefits as part of our compensation package to support our employees and their families. Our comprehensive benefits include global access to mental health and financial wellness support and resources. Local benefits include statutory and voluntary benefits which may include healthcare (medical, dental, and vision), life, accident, disability, commuter, and retirement options (401(k)/pension). Our employees can take time off work for vacation and other personal reasons to balance their evolving work and life needs. It's important to note that not every benefit is available in all locations or for every role. For details specific to your location, please consult with your recruiter.
The Roku Culture
Roku is a great place for people who want to work in a fast-paced environment where everyone is focused on the company's success rather than their own. We try to surround ourselves with people who are great at their jobs, who are easy to work with, and who keep their egos in check. We appreciate a sense of humor. We believe a fewer number of very talented folks can do more for less cost than a larger number of less talented teams. We’re independent thinkers with big ideas who act boldly, move fast and accomplish extraordinary things through collaboration and trust. In short, at Roku you'll be part of a company that's changing how the world watches TV.
We have a unique culture that we are proud of. We think of ourselves primarily as problem-solvers, which itself is a two-part idea. We come up with the solution, but the solution isn't real until it is built and delivered to the customer. That penchant for action gives us a pragmatic approach to innovation, one that has served us well since 2002.