machine learning engineer岗位职责

工程概览 ：
我们的目标是研发一款多功能的抓取与放置机器人，能够无缝集成至商用厨房中。 该机器人能识别多种原材料——从肉类到蔬菜——并依照用户指定的重量或数量进行量测，再将食材送至烹饪流程的下一阶段，全程不需人工干预。 此系统采用模块化的硬件与软件架构，可因应厨房布局变化与菜单更新调整。 通过强化学习驱动的运动策略，机器人能够在不同对象与工作空间间泛化其抓取与放置方式。
岗位简介：
我们正在寻找一位机器学习工程师，您将负责设计、开发与部署驱动机器人行为的学习策略，并在基于 ROS的框架中运行。 您的职责是实现感知系统与底层运动规划的无缝集成，即在实时ROS框架内，对图像、点云及力读数进行处理与转化，并传递给规划层，在实时 ROS 生态系统内实现端到端的性能与可靠性。
主要职责：
1、数据工程与前处理
1）主导数据流程：收集、标注及前处理多模态数据集，用于训练与评估
2）清理与标注数据，确保训练的完整性与可扩展性
2、算法开发与训练
1）在模拟与实际环境中开发并训练机器学习算法
2）定义奖励函数，监控训练指标，分析失败案例，并优化模型结构以提升策略鲁棒性
3、模型整合与部署
1）将训练好的模型整合到 ROS 控制回路中，优化边缘设备推理效能、延迟及吞吐量
2）评估系统性能，分析故障并对模型架构与奖励函数进行迭代改进
3）与跨学科/领域团队紧密合作，确保模型与机器人运动规划的无缝衔接
职位要求：
1.计算机科学、机器学习、机器人或相关领域学士（或以上）学位
2.精通 Python 及至少一种深度学习框架（如 PyTorch、TensorFlow）
3.具备强化学习和/或模仿学习实践经验
4.熟悉模拟工具（Gazebo、CoppeliaSim）与 GPU 加速训练流水技术与流程
加分条件：
1.具相关领域硕士学位
2.曾将ML模型整合至 ROS 或其他机器人中间件
3.掌握OpenCV与ROS的集成技术，并应用于机器人视觉开发
4.具有在自定义数据集上微调大型预训练视觉语言或多模态模型的经验
5.熟悉強化学习库（如 Ray R. Lilly， Stable Baselines）
6.熟悉推論优化框架（TensorRT、OpenVINO）
7.具备出色的问题解决能力，能于跨领域/学科团队中协作
Project Overview：
Our goal is to develop a versatile pick-and-place robot that seamlessly integrates commercial kitchens. It will identify a wide range of raw ingredients—ranging from meats to vegetables—measure them to user-specified weights counts, deposit them the next stage of the cooking line without human intervention. Built around modular hardware software components, the system will adapt to changing kitchen layouts evolving menus. By leveraging reinforcement-learning-driven–driven motion policies, the robot can generalise its grasping placement strategies across new objects workspaces with minimal reprogramming.
Job Summary：
We seek a proactive Machine Learning Engineer to design, develop, deploy machine learning policies that drive the robot’s behaviour within a ROS-based framework. You’ll bridge the gap between perception outputs—images, point clouds, force readings—low-level motion planning in a real-time ROS ecosystem. Close collaboration with controls, vision, systems teams will ensure end-to-end performance reliability.
Key Responsibilities：
1.Data Engineering & Preprocessing
1) Lead data-engineering pipelines: collect, annotate, preprocess multi-modal datasets ftraining evaluation.
2) Annotate, clean, preprocess datasets to ensure training integrity scalability
2.Algorithm development training
1) Develop train machine learning models in simulation the real world.
2) Define reward functions, monittraining metrics, analyse failure cases, iterate on model structures to improve policy robustness.
3.Model integration deployment
1)Integrate trained models ROS control loops, optimising fedge-device inference, latency, throughput.
2)Evaluate system performance, analyse failures, iterate on model architectures reward functions.
3)Work closely with a multidisciplinary team to ensure the model seamlessly integrates with robot motion planning.
Minimum Qualifications：
1.Bachelor’s degree (higher) in Computer Science, Machine Learning, Robotics a related field.
2.Proficiency in Python one more deep learning frameworks(e.g., PyTorch, TensorFlow).
3.Hands-on experience with reinforcement-learning and/imitation-learning workflows.
4.Familiarity with simulation tools (Gazebo, CoppeliaSim) GPU-accelerated training pipelines
Preferred Qualifications：
1.Master’s degree in a relevant discipline
2.Experience integrating ML models within ROS other robotic middleware
3.Experience with OpenCV ROS integration fvision-based robotics.
4.Experience fine-tuning large pretrained vision-language
multimodal models on custom datasets
5.Knowledge of RL libraries(Ray RLlib, Stable Baselines).
6.Hands-on expertise with inference optimisation frameworks(TensorRT, OpenVINO).
7.Strong problem-solving skills the ability to work collaboratively within a multidisciplinary team.

工程概览 ：
我们的目标是研发一款多功能的抓取与放置机器人，能够无缝集成至商用厨房中。 该机器人能识别多种原材料——从肉类到蔬菜——并依照用户指定的重量或数量进行量测，再将食材送至烹饪流程的下一阶段，全程不需人工干预。 此系统采用模块化的硬件与软件架构，可因应厨房布局变化与菜单更新调整。 通过强化学习驱动的运动策略，机器人能够在不同对象与工作空间间泛化其抓取与放置方式。
岗位简介：
我们正在寻找一位机器学习工程师，您将负责设计、开发与部署驱动机器人行为的学习策略，并在基于 ROS的框架中运行。 您的职责是实现感知系统与底层运动规划的无缝集成，即在实时ROS框架内，对图像、点云及力读数进行处理与转化，并传递给规划层，在实时 ROS 生态系统内实现端到端的性能与可靠性。
主要职责：
1、数据工程与前处理
1）主导数据流程：收集、标注及前处理多模态数据集，用于训练与评估
2）清理与标注数据，确保训练的完整性与可扩展性
2、算法开发与训练
1）在模拟与实际环境中开发并训练机器学习算法
2）定义奖励函数，监控训练指标，分析失败案例，并优化模型结构以提升策略鲁棒性
3、模型整合与部署
1）将训练好的模型整合到 ROS 控制回路中，优化边缘设备推理效能、延迟及吞吐量
2）评估系统性能，分析故障并对模型架构与奖励函数进行迭代改进
3）与跨学科/领域团队紧密合作，确保模型与机器人运动规划的无缝衔接
职位要求：
1.计算机科学、机器学习、机器人或相关领域学士（或以上）学位
2.精通 Python 及至少一种深度学习框架（如 PyTorch、TensorFlow）
3.具备强化学习和/或模仿学习实践经验
4.熟悉模拟工具（Gazebo、CoppeliaSim）与 GPU 加速训练流水技术与流程
加分条件：
1.具相关领域硕士学位
2.曾将ML模型整合至 ROS 或其他机器人中间件
3.掌握OpenCV与ROS的集成技术，并应用于机器人视觉开发
4.具有在自定义数据集上微调大型预训练视觉语言或多模态模型的经验
5.熟悉強化学习库（如 Ray R. Lilly， Stable Baselines）
6.熟悉推論优化框架（TensorRT、OpenVINO）
7.具备出色的问题解决能力，能于跨领域/学科团队中协作
Project Overview：
Our goal is to develop a versatile pick-and-place robot that seamlessly integrates commercial kitchens. It will identify a wide range of raw ingredients—ranging from meats to vegetables—measure them to user-specified weights counts, deposit them the next stage of the cooking line without human intervention. Built around modular hardware software components, the system will adapt to changing kitchen layouts evolving menus. By leveraging reinforcement-learning-driven–driven motion policies, the robot can generalise its grasping placement strategies across new objects workspaces with minimal reprogramming.
Job Summary：
We seek a proactive Machine Learning Engineer to design, develop, deploy machine learning policies that drive the robot’s behaviour within a ROS-based framework. You’ll bridge the gap between perception outputs—images, point clouds, force readings—low-level motion planning in a real-time ROS ecosystem. Close collaboration with controls, vision, systems teams will ensure end-to-end performance reliability.
Key Responsibilities：
1.Data Engineering & Preprocessing
1) Lead data-engineering pipelines: collect, annotate, preprocess multi-modal datasets ftraining evaluation.
2) Annotate, clean, preprocess datasets to ensure training integrity scalability
2.Algorithm development training
1) Develop train machine learning models in simulation the real world.
2) Define reward functions, monittraining metrics, analyse failure cases, iterate on model structures to improve policy robustness.
3.Model integration deployment
1)Integrate trained models ROS control loops, optimising fedge-device inference, latency, throughput.
2)Evaluate system performance, analyse failures, iterate on model architectures reward functions.
3)Work closely with a multidisciplinary team to ensure the model seamlessly integrates with robot motion planning.
Minimum Qualifications：
1.Bachelor’s degree (higher) in Computer Science, Machine Learning, Robotics a related field.
2.Proficiency in Python one more deep learning frameworks(e.g., PyTorch, TensorFlow).
3.Hands-on experience with reinforcement-learning and/imitation-learning workflows.
4.Familiarity with simulation tools (Gazebo, CoppeliaSim) GPU-accelerated training pipelines
Preferred Qualifications：
1.Master’s degree in a relevant discipline
2.Experience integrating ML models within ROS other robotic middleware
3.Experience with OpenCV ROS integration fvision-based robotics.
4.Experience fine-tuning large pretrained vision-language
multimodal models on custom datasets
5.Knowledge of RL libraries(Ray RLlib, Stable Baselines).
6.Hands-on expertise with inference optimisation frameworks(TensorRT, OpenVINO).
7.Strong problem-solving skills the ability to work collaboratively within a multidisciplinary team.

职责描述：
1.负责机器学习/深度学习模型的开发、训练、调优与部署；
2.处理海量数据，进行特征工程，优化模型性能指标（如准确率、召回率）；
3.搭建模型Pipeline，实现自动化训练与部署（与MLOps协作）；
4. 持续监控线上模型效果，迭代优化模型算法；
5.研究跟进前沿算法（如Transformer、GNN等）在业务场景的落地。
任职要求：
1.硕士及以上学历
2.-熟练掌握Python、PyTorch/TensorFlow、Scikit-learn、SQL/Hive、特征工程、模型评估、Linux、Git
（加分项：分布式训练框架如Horovod，云平台如AWS/Azure/GCP）