Smart cockpit AI model development_XC

Role Summary

Core model optimization: Leading the post-training and fine-tuning of cockpit core models such as speech recognition and multi-modal fusion, breaking through the difficulties of Code-Switching (mixed Chinese and English) and dialect-Mandarin mixed recognition, and improving the accuracy of command analysis.

Building interactive capabilities: Designing the cockpit Agent Function Calling framework, optimizing the precise mapping of natural language intentions to vehicle control APIs, and solving parameter filling and dependency reasoning for complex interactions such as multi-round dialogues and contextual reference.

Multi-modal fusion: Develop a four-modal fusion decision-making model of vision, voice, touch, and vehicle status signals to handle cross-modal conflict scenarios and optimize the efficiency of multi-modal collaborative decision-making.

Implementation and adaptation: Responsible for the collaborative adaptation of models and cockpit software and hardware, supporting product development, demo verification and auto show display needs, and leading the formulation of relevant technical specifications.

Requirements

• Education and experience: Bachelor degree or above, major in computer or AI, 3-5 years or more of post-training experience in AI models, preference will be given to those with experience in smart cockpit/vehicle AI projects.
• Technical capabilities: Proficient in TensorFlow/PyTorch, mastering the entire process of model fine-tuning, quantification, and distillation; familiar with speech recognition and multi-modal fusion algorithms, with practical experience in Code-Switching and multi-modal conflict handling.
• Practical ability: Ability to independently design post-training programs and lead core modules. Applicants with team guidance experience will be given priority; have strong problem analysis and technology implementation capabilities, and adapt to the R&D rhythm.