• Strategic Model Integration: Evaluation and integration of open-source and commercial models (e.g., for ASR, TTS, diarization, VAD), balancing quality, latency, and cost to accelerate game development. • Model R&D & Adaptation: Lead the fine-tuning and adaptation of speech models (TTS, Voice Cloning, ASR) using parameter-efficient techniques (e.g., LoRA/PEFT) to address production gaps and creative constraints that external/commercial solutions cannot satisfy. • Data & Alignment: Lead audio and multimodal data curation, including cleaning, segmentation, labeling, and synthetic data generation to support speech model training and adaptation. • Scalable Training: Scale speech model training using distributed data pipelines and training strategies for large datasets.