🌏 全球远程工作机会汇总

在招 0 仅中国地区远程 0 👁 累计浏览 275 更新时间

研究工程师(隐私与匿名化)

Research Engineer Privacy and Anonymization

全球可投 设计 预计有效期:2026-12-27

🏢 公司 / 雇主

Pragmatike

发布于 2026-11-12 · 首次收录 2026-11-16 · 最后确认在线 2026-11-17 · 👁 浏览 2 次

🛠 技能要求

Python机器学习数据匿名化信息提取命名实体识别(NER)分类模型大语言模型(LLM)统计模型差分隐私k-匿名安全聚合格式保留加密隐私增强技术(PETs)数据处理管道机器学习推理

📋 职位要求

  • 构建检测 PII、准标识符、凭证及其他敏感信息的系统,并根据数据类型和下游用例设计相应的转换方法。 Build systems to detect PII, quasi-identifiers, credentials and other sensitive information, and design transformations based on the data type and the downstream use case.
  • 开发并对检测方法进行基准测试,综合运用规则、统计模型、分类器和基于 LLM 的方法。 Develop and benchmark detection approaches that combine rules, statistical models, classifiers and LLM-based methods.
  • 构建生产级管道,在原始数据进入下游处理、训练、评估或合成数据生成之前对其进行匿名化。 Build production pipelines that anonymize raw data before it enters downstream processing, training, evaluation or synthetic data generation.
  • 创建评估框架,衡量隐私风险与保留的数据效用,包括召回率加权指标、泄露测试以及对抗性重识别尝试。 Create evaluation frameworks that measure privacy risk and retained data utility, including recall-weighted metrics, leakage tests and adversarial re-identification attempts.
  • 设计能够应对新数据源、模式漂移、异常格式以及意外字段中嵌入敏感信息的稳健系统。 Design systems that stay robust to new data sources, schema drift, unusual formats and sensitive information embedded in unexpected fields.
  • 精通 Python,并有构建可靠的生产级数据或机器学习系统的经验。 Strong proficiency in Python, and experience building reliable production data or machine-learning systems.
  • 具备信息提取、命名实体识别、分类或相关方法的经验,用于检测罕见或敏感内容。 Experience with information extraction, named-entity recognition, classification, or related methods for detecting rare or sensitive content.
  • 拥有出色的实验直觉,能够在召回率、精确率、延迟、成本和下游数据效用等维度上比较不同方法。 Strong experimental instincts, and the ability to compare approaches across recall, precision, latency, cost and downstream data utility.
  • 理解删除、掩码、假名化、匿名化和合成数据之间的区别,以及各自适用的场景。 An understanding of the difference between redaction, masking, pseudonymization, anonymization and synthetic data — and when each is appropriate.
  • 高度注重细节,能够推演隐蔽的泄露路径、边缘情况和对抗性失败模式。 High attention to detail, and the ability to reason about subtle leakage paths, edge cases and adversarial failure modes.
  • 具备端到端构建数据处理管道的经验,包括在没有完整既定路线图的情况下开展工作。 Experience building data-processing pipelines end to end, including where there was no fully prescribed roadmap.
  • 具备清晰的英语书面和口头表达能力,能够跨团队、跨时区顺畅协作。 Clear written and spoken English, and comfort collaborating across teams and time zones.

📄 职位描述(中文)

地点:旧金山或新加坡(现场办公),或远程(新加坡或美国太平洋时区) 入职时间:尽快(ASAP) 语言:英语 团队:工程 关于该职位 我们的客户是一家快速发展的 AI 公司,致力于为 AI 训练数据构建可靠、公平且开放的基础设施。其市场平台让数据创作者能够从数据中获得价值,并为 AI 实验室提供可信赖的训练数据。该平台已被初创公司、财富 500 强企业和前沿实验室采用,公司获得了包括 Y Combinator 在内的顶尖风险投资机构的支持。为满足强劲需求,公司正在快速扩张。 我们正在寻找一位研究工程师(隐私与匿名化),负责构建相关系统,使敏感的现实世界数据变得安全、同时仍对 AI 训练有用。你将开发各种方法,在原始数据进入处理流程和合成数据管道之前,检测并移除个人身份信息(PII)、机密凭证及其他敏感内容——在保护隐私的同时,不破坏使数据对智能体训练具有价值的结构和信号。 这是一个隐私与数据保护工程岗位,而非合成任务设计岗位。你将全面负责位于下游处理、训练、评估和合成生成之前的管道。 这是一份全职工作,尽快入职。办公室位于旧金山和新加坡;如果远程候选人的工作时间能与旧金山(美国太平洋时区)或新加坡的工作时间重叠约 70–80%,我们同样欢迎远程候选人。具体每日工作安排将在招聘过程中协商确定。 你的职责 构建检测 PII、准标识符、凭证及其他敏感信息的系统,并根据数据类型和下游用例设计相应的转换方法。 开发并对检测方法进行基准测试,综合运用规则、统计模型、分类器和基于 LLM 的方法。 构建生产级管道,在原始数据进入下游处理、训练、评估或合成数据生成之前对其进行匿名化。 创建评估框架,衡量隐私风险与保留的数据效用,包括召回率加权指标、泄露测试以及对抗性重识别尝试。 设计能够应对新数据源、模式漂移、异常格式以及意外字段中嵌入敏感信息的稳健系统。 与工程、研究、运营和客户团队合作,将隐私需求转化为切实可行的技术策略与保障措施。 我们的要求 精通 Python,并有构建可靠的生产级数据或机器学习系统的经验。 具备信息提取、命名实体识别、分类或相关方法的经验,用于检测罕见或敏感内容。 拥有出色的实验直觉,能够在召回率、精确率、延迟、成本和下游数据效用等维度上比较不同方法。 理解删除、掩码、假名化、匿名化和合成数据之间的区别,以及各自适用的场景。 高度注重细节,能够推演隐蔽的泄露路径、边缘情况和对抗性失败模式。 具备端到端构建数据处理管道的经验,包括在没有完整既定路线图的情况下开展工作。 具备清晰的英语书面和口头表达能力,能够跨团队、跨时区顺畅协作。 优秀的候选人还可能具备隐私增强技术(如差分隐私、k-匿名、安全聚合或格式保留加密)的实践经验;曾在医疗、金融或安全领域处理敏感数据;构建过低延迟或高吞吐量的机器学习推理或数据处理系统;或将工作从早期研究推进到生产落地。 技术能力与学习潜力比工作年限更重要。即使不完全满足上述每一条,也鼓励有积极性的候选人申请。 这个职位对你的意义 让现实世界的数据安全地用于训练:你将构建相关系统,使敏感数据可用于 AI 训练,同时不暴露人员或机密。 保护隐私但不损害效用:这项工作不只是删除数据,还要保留下游训练和合成管道所需的结构和信号。 从检测到生产:你将负责完整链路:检测方法、转换、评估,以及位于训练和生成环节之前的管道。 高难度、高风险的边缘情况:你将处理泄露路径、模式漂移和对抗性失败模式,而不仅仅是干净的基准数据集。 旧金山、新加坡或远程:可在旧金山或新加坡现场办公,或远程工作,工作时间与美国太平洋时区或新加坡工作时间重叠约 70–80%。具体每日工作安排将在招聘过程中协商确定。 Pragmatike 是平等机会雇主,致力于为所有申请人提供无歧视的平等就业机会。我们代表客户进行招聘,禁止基于种族、肤色、宗教、年龄、性别、原国籍、残疾状况、遗传信息、受保护退伍军人身份、性取向、性别认同或表达,以及受联邦、州或地方法律保护的任何其他特征的歧视和骚扰。本政策适用于雇佣的所有条款和条件,包括招聘、录用、安置、晋升、解雇、裁员、召回、调岗、休假、薪酬和培训。我们致力于打造公平、包容的招聘流程。我们仅出于招聘目的处理您的个人数据,遵守适用的隐私法律,并采取合理的保障措施保护您的信息。您的数据可能会与我们的客户共享以供录用评估,但不会向招聘流程之外的第三方披露。申请时请提及单词 **CONFIDENT** 并附上标签 RMjQwZTo0NTc6ZmYxMjozOWQ6NmQ0Nzo4MGE2Ojc1NzM6NjIzMw==,以表明您已完整阅读本招聘启事(#RMjQwZTo0NTc
🇬🇧 查看英文原文(Original description)
Location: San Francisco or Singapore (on-site), or remote (Singapore or US Pacific time zones) Start Date: ASAP Languages: English Team: Engineering About the Role Our client is a fast-growing AI company building reliable, fair and open infrastructure for AI training data. Their marketplace helps the people who create data get value for it, and gives AI labs trustworthy data to train on. It is already used by startups, Fortune 500 companies and frontier labs, and the company is backed by leading venture investors, including Y Combinator. They are scaling quickly to meet strong demand. We are looking for a Research Engineer, Privacy and Anonymization to build the systems that make sensitive, real-world data safe and still useful for AI training. You will develop methods to detect and remove personally identifiable information, secrets and other sensitive content from raw data before it enters processing and synthetic data pipelines — protecting privacy without destroying the structure and signal that make the data valuable for training agents. This is a privacy and data-protection engineering seat, not a synthetic-task design role. You will own the pipeline that stands in front of downstream processing, training, evaluation and synthetic generation. This is a full-time role, starting ASAP. Offices are in San Francisco and Singapore; remote candidates are welcome if their working hours can overlap about 70–80% with either San Francisco (US Pacific) or Singapore business hours. The exact daily schedule will be agreed during hiring. What You'll Do Build systems to detect PII, quasi-identifiers, credentials and other sensitive information, and design transformations based on the data type and the downstream use case. Develop and benchmark detection approaches that combine rules, statistical models, classifiers and LLM-based methods. Build production pipelines that anonymize raw data before it enters downstream processing, training, evaluation or synthetic data generation. Create evaluation frameworks that measure privacy risk and retained data utility, including recall-weighted metrics, leakage tests and adversarial re-identification attempts. Design systems that stay robust to new data sources, schema drift, unusual formats and sensitive information embedded in unexpected fields. Work with engineering, research, operations and customers to translate privacy requirements into practical technical policies and safeguards. What We're Looking For Strong proficiency in Python, and experience building reliable production data or machine-learning systems. Experience with information extraction, named-entity recognition, classification, or related methods for detecting rare or sensitive content. Strong experimental instincts, and the ability to compare approaches across recall, precision, latency, cost and downstream data utility. An understanding of the difference between redaction, masking, pseudonymization, anonymization and synthetic data — and when each is appropriate. High attention to detail, and the ability to reason about subtle leakage paths, edge cases and adversarial failure modes. Experience building data-processing pipelines end to end, including where there was no fully prescribed roadmap. Clear written and spoken English, and comfort collaborating across teams and time zones. Strong candidates may also have hands-on experience with privacy-enhancing technologies (such as differential privacy, k-anonymity, secure aggregation or format-preserving encryption); have worked with sensitive data in healthcare, finance or security; have built low-latency or high-throughput ML inference or data-processing systems; or have taken work from early research through to production. Technical aptitude and learning potential matter more than years of experience. Motivated candidates are encouraged to apply even if they don't meet every point above. Why This Role Matters for You Make real-world data safe to train on: You will build the systems that let sensitive data be used for AI training without exposing people or secrets. Privacy without wrecking utility: The job is not only to redact, but to keep the structure and signal that downstream training and synthetic pipelines need. From detection to production: You will own the full path: detection methods, transformations, evaluation, and the pipeline that sits in front of training and generation. Hard, high-stakes edge cases: Work on leakage paths, schema drift and adversarial failure modes, not only clean benchmark datasets. San Francisco, Singapore, or remote: Work on-site in San Francisco or Singapore, or remotely, with hours that overlap about 70–80% with either US Pacific or Singapore business hours. The exact daily schedule will be agreed during hiring. Pragmatike is an Equal Opportunity Employer and is committed to providing equal employment opportunities to all applicants without discrimination. We recruit on behalf of our clients and prohibit discrimination and harassment based on race, color, religion, age, sex, national origin, disability status, genetics, protected veteran status, sexual orientation, gender identity or expression, or any other characteristic protected by federal, state, or local laws. This policy applies to all terms and conditions of employment, including recruiting, hiring, placement, promotion, termination, layoff, recall, transfer, leaves of absence, compensation, and training. We are committed to a fair and inclusive hiring process. We process your personal data solely for recruitment purposes, in accordance with applicable privacy laws, and maintain reasonable safeguards to protect your information. Your data may be shared with our client(s) for hiring consideration, but will not be disclosed to third parties outside of the recruitment process. Please mention the word **CONFIDENT** and tag RMjQwZTo0NTc6ZmYxMjozOWQ6NmQ0Nzo4MGE2Ojc1NzM6NjIzMw== when applying to show you read the job post completely (#RMjQwZTo0NTc6ZmYxMjozOWQ6NmQ0Nzo4MGE2Ojc1NzM6NjIzMw==). This is a beta feature to avoid spam applicants. Companies can search these words to find applicants that read this and see they're human.