岗位描述
Open Position We are looking for a motivated student to join a paid, onsite 6-month Master’s thesis project starting in July at XPENG European R&D Center in Munich, focusing on generative AI. The student will become part of the AI Model Team and work under the supervision of experts in NLP and Audio Machine Learning. The project explores how generative audio methods can be used to create realistic background noise and spatial acoustic conditions for data augmentation and controlled evaluation. This opportunity is well suited for students with a strong interest in speech processing, audio machine learning, and generative AI. Overview and Motivation Modern multimodal voice assistants rely on large, high-quality speech datasets recorded in diverse acoustic environments. However, collecting and curating such real-world data is both costly and time-consuming, and often fails to sufficiently cover important edge cases and environmental variations. Recent advances in generative AI for audio have enabled the synthesis of realistic and controllable acoustic environments, including background noise and spatial sound characteristics. This thesis investigates whether these generative methods can be leveraged to augment or partially replace traditional data collection approaches. The goal is to evaluate whether synthetic acoustic environments can improve model robustness and downstream performance in controlled evaluation settings, thereby enhancing the overall quality of multimodal voice assistants.