ROBOTERA

大模型部署和加速工程师

ROBOTERA  •  Onsite  •  3 months ago
Apply
AI can make mistakes so check important info. Chat history is never stored.

Job Description

大模型部署和加速工程师北京全职职位描述1.参与针对大模型的训练推理加速、模型服务化、GPU资源调度等功能的研发;
2.模型轻量化:包括但不限于量化、剪枝、蒸馏、知识蒸馏、编译优化等,以满足端侧或云端部署的性能要求(低延迟、高吞吐、低功耗);
3.负责大模型封装、部署pipeline的开发,负责将优化后的模型部署到各种目标硬件平台(AGX、thor、4090等);职位要求1计算机及相关专业硕士及以上,2年以上工作经历;
2.扎实的计算机和软件工程基础,精通 C++ 、python编程,熟悉shell、docker使用,了解基本的操作系统、计算机体系结构、编程语言、计算机网络知识;
3.熟悉至少一种主流AI训练框架,如PyTorch、FensorFlow、FLAX等;
4.熟悉至少一种大模型加速技术,如kv-cache、flash-attention、trnsorrt-llm等,熟悉常见的大模型推理框架,如如vLLM、TGI、SGLang、TensorRT-LLM、LightLLM、lmdeploy等;
5.深入理解模型性能分析工具和方法,能够准确评估模型在不同硬件上的性能,熟悉GPU、NPU等硬件上的性能分析工具;
符合以下一项或多项条件者优先考虑:
1.端侧硬件实战经验:在端侧硬件(AGX、thor等)有实际加速深度学习模型经验,熟悉常见硬件的体系架构和性能特性,熟悉相关推理框架(如 TensorRT, ONNX Runtime, OpenVINO, TVM 等);
2.具备大模型训练、推理加速实战经验; 投递
ROBOTERA

About ROBOTERA

ROBOTERA is a leading robotics company, founded in 2023, specializing in fully in-house humanoid robots and dexterous hands. From core actuators and joints to complete robotic systems, we develop high-performance, reliable robots entirely in-house, ensuring precise control and robust engineering across all our products.

Our lineup includes the L7 full-size bipedal humanoid, built for complex motion and stability; the Q5 intelligent service robot, designed for service and research applications; and the XHAND1 dexterous hand, demonstrating advanced manipulation capabilities. Each product showcases our engineering excellence and hardware innovation.

By combining robust engineering with versatile robotic systems, ROBOTERA delivers solutions ready for real-world environments. Our mission is to advance robotics technology and provide reliable robots that enhance human productivity, safety, and collaboration across industries and research.

Industry
Manufacturing & Production
Company Size
11-50 employees
Headquarters
Hong Kong, HK
Year Founded
2023
Social Media