Senior Staff Research Engineer, Speech Machine Learning

Lab Summary:

Bixby is an intelligent personal assistant which is only available as a built-in application on Samsung flagship devices and wearables. This application uses Natural Language Understanding to perform tasks on these devices using voice/ text, including but not limited to making phone calls, sending text messages, setting up meetings, opening apps, setting alarms and timers, getting directions, answering general questions, providing information about restaurants and other businesses, etc.

Position Summary:

For this position we are expanding our AILs voice technology and features to include advanced research and projects in Wake word detection, Automatic Speech Recognition (ASR), that includes Acoustic and Language Modeling, and personalization. We also work on language and gender detection using speech signals, Speaker identification, verification and diarization techniques. At AIL we perform state-of-the-art research in multi-lingual/accents research and bringing those research ideas to production. We are looking for candidates with extensive expertise in Digital Signal/Speech Processing with Speech recognition specialization, demonstrated research expertise by publishing papers in reputed journals/conferences, excellent knowledge of  Deep/Machine Learning with 7+ years of industry experience. Candidates are expected to work in a fast paced environments. 

Position Responsibilities: 

• Architect and design end to end Automatic Speech Recognition products, applications and solutions for specific business needs and provide implementation guidance during delivery
• Leverage, customize and implement ASR models, algorithms, and methodologies to improve the overall quality ASR in various applications and systems
• Analyze and evaluate the performance ASR systems and provide design recommendations
• Analyze and make right technological choices for generative ai solutions
• Design and prototype reusable components for LLM based solutions for ASR
• Architect components of an ASR solution to address Responsible AI & Security
• Collaborate seamlessly with diverse, cross-functional teams to accurately identify and prioritize requirements, ensuring that the language model meets the needs and expectations of various stakeholders
• Create and maintain comprehensive technical documentation that comprehensibly captures the intricate details of the language model, facilitating seamless understanding, efficient troubleshooting, and future development
• Harness the power of transformer architecture, a cutting-edge deep learning model widely employed in natural language processing and computer vision, to optimize the language model's performance and efficiency
• Exploiting the transformative capabilities of transformer architectures to seamlessly process and reshape vast volumes of data, empowering the language model to achieve unprecedented levels of accuracy and versatility
• Ensure ethical AI development practices, prioritizing fairness, transparency, and privacy

Required Skills:

• MS or Ph.D. in Computer Science or Digital Signal Processing or equivalent combination of education, training, and experience
• 7+ years of relevant professional experience in Machine Learning or relevant field
• Experience with Tensorflow or Pytorch or similar frameworks
• Worked on advance architectures such as transformers, conformer and other advanced models for ASR systems
• Working experience on ASR in large scale production systems
• Experience in modeling ML algorithms on GPUs at scale
• Experience with multi-lingual speech, low resource speech research and architectures
• Working experience on deploying recognition engines on both server and edge devices
• Experience with Acoustic modeling, noise and ambient modeling, and its effects on ASR
• Knowledge of state-of-the-art Large Language models such as Deepseek, GPT, BERT variants and other deep fusion techniques is essential
• Working on WFST, n-gram and other shallow fusion techniques for named entity recognitions
• Experience on speaker recognition, wakeup and audio-based language recognition is desirable
• Experience with improving ASR performance in far field and noisy environments
• Working experience on masking and spectral restoration based noise suppression and speech enhancement techniques
• Experience in developing advance classification models such as ECAPA-TDNN for speaker, gender classifications
• Ability to develop project plans and experience to execute them
• Research expertise in ML and written research publications
• C/C++, PYTHON, JAVA programming language experience
• Leadership ability to lead a mid-size team

Our total