Description and Requirements
1. On-Device Engineering & Deployment: Turn the team's frontier research (e.g., diffusion language models, Jev-like structured decision models, latent-space reasoning models) from algorithm prototypes into engineering solutions that run reliably on device. Own model adaptation, performance optimization, and integration delivery.
2. Hardware-Software Co-Optimization: Work with chip and system teams on heterogeneous on-device hardware (NPU/GPU/CPU), covering inference engine adaptation, operator development and optimization, quantization and compression, and memory and power optimization, so that solutions meet on-device constraints on latency, power, and resources.
3. Training & Inference Infrastructure: Build and maintain the training and inference engineering stack that supports research on emerging paradigms, including distributed training frameworks, experiment management, and evaluation pipelines, to speed up the research team's iteration.
4. Project Management & Delivery: Manage projects end to end, from kickoff to delivery. Set plans and milestones, track progress, and identify risks, so that research outcomes become on-device product capabilities or solutions on time and to quality standards.
5. Cross-Team Collaboration: Serve as the bridge between the research team and engineering, product, and hardware teams. Align requirements and technical solutions, and drive cross-team issues to closure.
6. Engineering Standards & Best Practices: Establish model delivery standards, performance benchmarks, and engineering guidelines, and build reusable toolchains and best practices.


