机器学习与应用研究 · 中国 · 深圳Machine Learning & Applied Research · Shenzhen, China
机器学习(ML)工程师 / 科学家Machine Learning (ML) Engineer / Scientist
把数据、实验与模型转化为可验证、可部署的机器学习能力。Turn data, experiments and models into validated, deployable machine learning capabilities.
- 工作地区Market
- 中国China
- 所属团队Team
- 机器学习与应用研究Machine Learning & Applied Research
- 工作地点Working location
- 中国 · 深圳Shenzhen, China
- 聘用安排Employment terms
- 具体条件与团队沟通确认To be discussed with our team
职位简介The role
你的工作将带来什么价值。Where you’ll make a difference.
加入深圳团队,将应用研究转化为可实际使用的模型能力。围绕明确的客户问题,参与数据准备、实验、评测与推理集成。我们重视可复现的证据、细致的误差分析和工程落地能力,而不仅是模型的新颖性。Join the Shenzhen team to turn applied research into usable model capabilities. Work across data preparation, experimentation, evaluation and inference integration for defined client problems. We value reproducible evidence, thoughtful error analysis and practical engineering as much as model novelty.
工作职责Responsibilities
你将负责的工作。What you’ll own.
- 01
评估数据质量,建立基线、实验方案和符合场景的验证指标。Assess data quality and establish baselines, experiments and scenario-specific metrics.
- 02
训练、微调或适配模型,分析误差、偏差及边界情况。Train, fine-tune or adapt models; analyse errors, bias and edge cases.
- 03
协作完成推理接口、性能优化、版本管理和上线监测。Collaborate on inference APIs, performance, versioning and production monitoring.
- 04
检查数据来源、标注和质量,设计避免数据泄漏且贴近部署场景的训练、验证与测试集划分。Audit dataset provenance, labelling and quality. Design train/validation/test splits that avoid leakage and reflect the intended deployment.
- 05
使用版本化数据、配置和基线开展受控实验,完整记录准确性、鲁棒性和计算成本,不只挑选表现良好的样例。Run controlled experiments with versioned datasets, configurations and baselines. Record accuracy, robustness and compute cost rather than selecting only favourable examples.
- 06
分析不同失败场景,改进模型或数据策略;与应用工程师协作优化推理接口、批处理、延迟及资源使用。Analyse failure slices and improve model or data choices. Work with application engineers on inference APIs, batching, latency and resource usage.
- 07
记录模型限制、评测结果与复现步骤,并与交付团队定义上线监控、重新训练和回滚的触发条件。Document model limitations, evaluation results and reproducibility steps. Define production monitoring and retraining or rollback triggers with the delivery team.
任职要求Requirements
我们期待的能力。What you’ll bring.
- 经验要求Experience
- 至少 3 年相关机器学习工程或应用研究经验;有可复现实验和可运行原型支持的实质性研究生研究经历可计入。At least 3 years of relevant ML engineering or applied research. Substantial postgraduate research may count when supported by reproducible experiments and deployed prototypes.
- 学历与资历Qualifications
- 计算机、数学、统计或相关专业本科,或同等机器学习实践经验。研究方向优先考虑相关硕士或博士,工程方向不强制要求研究生学历。Degree in computer science, mathematics, statistics or a related field, or equivalent applied ML experience. A relevant master’s or PhD is preferred for research-focused work, not mandatory for engineering candidates.
- 工具与技术知识Tools & technical knowledge
- 掌握 Python、NumPy/pandas、scikit-learn 及 PyTorch 或 TensorFlow;熟悉 Git、SQL 和 MLflow 等实验追踪工具,理解数据划分、数据泄漏、评估指标、推理及可复现性。Python, NumPy/pandas, scikit-learn and PyTorch or TensorFlow; Git, SQL and experiment tracking such as MLflow. Understand data splits, leakage, metrics, inference and reproducibility.
- 作品与案例What to show us
- 提供两个实验或模型案例,包括基线、数据集说明、评测结果和误差分析;如有部署或推理性能结果请一并说明,并保护客户数据。Share two experiments or models with a baseline, dataset description, evaluation results and error analysis. Include deployment or inference performance where available; protect client data.
- 01
掌握 Python、机器学习基础及常用训练和数据处理工具。Strong Python and machine learning foundations, including training and data-processing tools.
- 02
具备模型实验、评估或部署经验,能说明个人贡献和实验结论。Model experimentation, evaluation or deployment experience, with clear personal contributions and findings.
- 03
能够记录实验过程,清晰沟通模型局限、数据问题及工程取舍。Ability to document experiments and communicate model limitations, data issues and engineering trade-offs.
- 04
理解统计验证、过拟合、类别不平衡和指标选择,能区分真实改进与随机波动。Understand statistical validation, overfitting, class imbalance and metric selection. Be able to distinguish a meaningful improvement from noise.
- 05
能编写可读、可测试的 Python,并使用中文及工作英语解释实验。研究方向需展示可复现代码,工程方向需说明部署约束。Write tested, readable Python and explain experiments in Mandarin and working English. Research candidates should demonstrate reproducible code; engineering candidates should explain deployment constraints.
加分经验Nice to have
有助于开展工作的经验。Useful additional experience.
- 有计算机视觉、自然语言处理、多模态系统或 MLOps 经验。Experience with computer vision, natural language processing, multimodal systems or MLOps.
- 具备 ONNX、模型量化、GPU 性能分析或容器化推理服务经验。Experience with ONNX, model quantisation, GPU profiling or containerised inference services.
- 具备合成数据评估、标注流程或多模态误差分析经验。Experience with synthetic data evaluation, annotation workflows or multimodal error analysis.
协作方式Ways of working
与团队一起推进工作。How you’ll work with the team.
- 01
让实验可复现,让结论有证据。Make experiments reproducible and conclusions evidence-based.
- 02
与产品和工程团队共同定义成功标准及监控需求。Define success and monitoring needs with product and engineering teams.
休假与法定福利Leave & statutory benefits
适用于中国大陆劳动关系。福利按国家及用工所在地规定执行,并结合累计工龄和个人参保资格确定;具体安排在劳动合同中明确。For employment in mainland China. Benefits follow national law and the rules of the employing city, taking service history and individual eligibility into account. The employment agreement will specify applicable terms.
带薪年休假Paid annual leave
连续工作满 1 年后,累计工龄满 1 年不满 10 年为 5 天,满 10 年不满 20 年为 10 天,满 20 年为 15 天;法定不享受情形及折算规则适用。After at least one year of continuous work: 5 days for cumulative service of 1 to under 10 years, 10 days for 10 to under 20 years, and 15 days from 20 years, subject to statutory exclusions and pro-rating.
法定节假日与休息Public holidays & rest
依法享有法定节假日及相应工时制度下的休息安排;加班工资或补休依适用法律执行。Statutory public holidays and rest arrangements under the applicable working-time system. Overtime pay or compensatory rest follows the relevant legal rules.
社会保险与住房公积金Social insurance & housing fund
按所在地规定及个人资格缴纳法定社会保险和住房公积金,缴费基数与比例以用工所在地规定为准。Statutory social insurance and housing provident fund contributions according to local rules and individual eligibility; contribution bases and rates depend on the employing location.
医疗与家庭相关假期Medical & family leave
依法安排医疗期、病假待遇及符合条件的产假等家庭相关假期,具体标准按国家和当地规定执行。Statutory medical-treatment periods and sick-pay arrangements, plus maternity and other eligible family leave, follow national and local rules.
Official references / 官方依据
申请意向Express interest
让我们了解你的经验与作品。Show us the work behind your experience.
请发送简历、说明岗位匹配度的简短自我介绍,以及相关作品、案例或技术项目链接。注明当前所在地、通知期或最早到岗时间,说明每个案例中的个人贡献,并删除客户机密信息。请注明职位编号:Send your résumé, a short introduction explaining why this role fits you, and relevant portfolio, case-study or technical project links. Tell us your current location and notice period or earliest availability. Explain your own contribution in each example and remove confidential client information. Include this role reference: CN-02
提交此职位的申请意向Express interest in this role
