Embodied-AI & Human-Robot Interaction Research · Remote 具身智能与人机交互研究 · 远程弹性工作制
- RynnBrain VLM × OmniGibson evaluation pipeline: Lifted BDDL true task success from 0% to 80% by replacing six classes of privileged simulator state with real perception and motion planning (GT segmentation → GSAM open-vocabulary perception; container AABBs → pixel-level snap; teleportation → CuRobo trajectory planning) across 60+ seeds and 3 BEHAVIOR-1K tasks, documenting 28 falsified hypotheses. RynnBrain VLM × OmniGibson 评测管线：剥离 6 类仿真特权信息（真值分割替换为 GSAM 开放词汇感知、容器包围盒替换为像素级对齐、瞬移替换为 CuRobo 轨迹规划），在 3 项 BEHAVIOR-1K 任务与 60+ 随机种子上将 BDDL 真实成功率从 0% 提升至 80%，完整归档 28 项证伪假设。
- Humanoid voice interaction: Co-deployed a real-time domestic voice stack on a humanoid robot for water-fetch and reception scenarios using Alibaba Cloud TTS/STT, Agora RTC, and TEN Framework, researching 3C noise suppression and microphone directivity. 人形机器人实时语音交互：在人形机器人上联合部署迎宾与递水场景的全栈国产化实时语音系统（阿里云语音合成/识别、声网 RTC、TEN Framework），研究软硬件边界的 3C 降噪与麦克风指向性。
- Multi-cloud GPU infrastructure: Collaborated with several partner infrastructure teams on GPU container scheduling, heterogeneous domestic/NVIDIA cluster orchestration, and vLLM inference pipelines. 多云异构算力协同：与多家合作方基础设施团队协同排查 GPU 容器调度、国产芯片与 NVIDIA 混合集群编排及共享算力推理管线。