（一）岗位概述

English: You will help build the evaluation system and testing pipelines for the company’s AI inspection business. You will conduct tests and evaluations for AI algorithms including NLP, CV and LLM to build trustworthy and compliant AI detection capabilities. This role mainly targets professionals with 1–5 years of working experience. Outstanding fresh graduates and interns are also welcome to apply. 中文：你将参与搭建公司 AI 检测业务的评测体系与测试流程，面向 NLP、CV、LLM 等 AI 算法开展测试与评测，构建可信、合规的 AI 检测能力。本岗位以 1–5 年社招人员为主，同时欢迎优秀应届生与实习生投递。

（二）岗位职责

English: Job responsibilities are divided into three tiers corresponding to different experience levels and growth paths. 中文：岗位职责分为三个层级，分别对应不同经验阶段候选人的工作重心与成长路径。

English: Core Execution Tier (1–5 years of working experience): Design, write and execute AI test cases, and deliver test and evaluation reports. Help build and optimize AI evaluation workflows, construct and maintain evaluation corpora and datasets. Evaluate LLMs from functionality, performance, content safety and other dimensions, deploy and use security toolchains such as LlamaGuard and ShieldLM. Take part in red-team adversarial testing and result analysis, and assist in drafting technical documents for detection standards and testing methods. Keep track of cutting-edge AI evaluation methods and tools to continuously improve evaluation efficiency and coverage. 中文：核心执行层（社招 1–5 年）：负责 AI 相关测试用例的设计、编写与执行，输出测试 / 评测报告。参与 AI 评测流程的搭建与优化，构建、维护评测语料库与评测数据集。执行大模型功能、性能、内容安全等维度的评测，部署与使用安全检测工具链（如 LlamaGuard、ShieldLM 等）。参与红队对抗测试的执行与结果分析，协助完成检测标准与测试方法的技术文档编写。跟踪 AI 评测前沿方法与工具，持续优化评测效率与覆盖度。

English: Advanced Expansion Tier (Senior Members): Participate in the construction of the AI evaluation system covering LLM content safety evaluation, red-team adversarial testing, compliance certification assessment and other areas, and promote the standardization of evaluation methodologies. Support communication with clients and business teams, understand requirements, deliver detection and certification solutions, and control project quality and delivery. Follow updates on AI regulations and compliance requirements such as the EU AI Act and minor protection clauses, and align evaluation standards with relevant laws and regulations. 中文：进阶拓展层（资深者）：参与 AI 评测体系建设，覆盖大模型内容安全评测、红队对抗测试、合规认证测评等方向，推动测评方法学标准化。协助对接客户与业务方，理解需求并输出检测 / 认证解决方案，把控项目质量与交付。跟踪 AI 监管与合规动态（如欧盟 AI Act、未成年人保护条款等），推动测评标准与法规对齐。

English: Foundational Learning Tier (Fresh Graduates / Interns): Assist in building, cleaning and annotating AI evaluation corpora and datasets. Support evaluation tasks, sort out and analyze evaluation results, and help draft test reports and technical documents. Research AI evaluation methods and security detection tools and produce research notes. Learn and master the AI evaluation framework under mentor guidance. 中文：培养起步层（校招 / 实习生）：协助构建、清洗与标注 AI 评测语料库 / 数据集。协助执行评测任务并整理分析评测结果，协助撰写测试报告与技术文档。调研 AI 评测方法与安全检测工具，输出调研笔记；在导师指导下学习并掌握 AI 评测方法体系。

（三）任职要求

English: Requirements are set for three groups: experienced hires, outstanding fresh graduates and interns. Candidates satisfying any set of requirements may apply. 中文：任职要求按社招、优秀校招、实习生三档分别列出，满足其中一档即可投递。

English: Experienced Hires (1–5 years): Bachelor’s degree or above, preferably in Computer Science, AI, Mathematics or related fields. 1–5 years of experience in software testing, development or project management; exceptions may be made for exceptional fresh graduates. In-depth understanding of NLP, CV, LLM and relevant AI testing approaches. Familiar with test case design, assessment and testing pipeline setup. Proficient in Python or at least one programming language; experience in automated testing and evaluation scripting is preferred. Passionate about AI evaluation, proactive, good at communication and teamwork, strong logical thinking and willing to pursue continuous improvement. 中文：社招 1–5 年・主要对象：本科及以上学历，计算机 / AI / 数学等相关专业优先。1–5 年软件测试 / 开发 / 项目管理相关经验（优秀应届生可放宽）。深入理解 NLP/CV/LLM 等 AI 算法及其测试方法。熟悉测试用例的设计、评估与测试流程搭建。掌握 Python 等至少一门编程语言，有自动化测试 / 评测脚本经验者优先。对 AI 评测有热情，态度积极，具备沟通与协作能力；逻辑思维强，能持续改进工作。

English: Outstanding Fresh Graduates (Class of 2027 or 2026; within 1 year after graduation is acceptable): Bachelor’s degree or above for graduates of 2027 or 2026, majoring in Computer Science, Artificial Intelligence, Software Engineering, Mathematics or related disciplines. Solid computer fundamentals, proficient in at least one programming language (Python preferred). Basic knowledge of NLP, CV and LLM, enthusiasm for generative AI and LLM evaluation. Fast learner, logical and good at communication and teamwork; relevant internship, project or competition experience is a plus. 中文：优秀校招放宽（2027 届或 2026 届・毕业 1 年内亦可）：2027 届或 2026 届本科及以上应届毕业生，计算机 / 人工智能 / 软件工程 / 数学等相关专业。具备扎实的计算机基础，掌握至少一门编程语言（Python 优先）。了解 NLP/CV/LLM 等 AI 技术基础，对生成式 AI / 大模型评测有热情。学习能力强、逻辑清晰，善于沟通与团队协作；有相关实习、项目或竞赛经验者优先。

English: Interns (Current Students): Bachelor’s or master’s students majoring in Computer Science, Artificial Intelligence, Software Engineering, Mathematics or related majors. Basic programming skills in Python and an interest in LLM evaluation and AI security. Able to work no less than 4 days per week; internships lasting 5 months or longer are preferred. Meticulous, self-motivated and good at communication. 中文：实习生放宽・在校生・每周 4 天以上：在校本科 / 硕士生，计算机、人工智能、软件工程、数学等相关专业。掌握 Python 等编程基础，对大模型评测、AI 安全方向感兴趣。每周可实习 4 天以上、实习期 5 个月以上者优先。认真细致，主动好学，具备良好的沟通能力。

（四）加分项（非必须）

English: Bonus qualifications (not mandatory): LLM safety evaluation (content safety, jailbreak, bias), red-team adversarial testing (garak, PyRIT, etc.), Python automated evaluation scripts, content safety detection models (LlamaGuard, ShieldLM, etc.), AI compliance and certification experience, alignment with evaluation benchmarks (e.g. AILuminate), corpus construction and data annotation, NLP and machine learning fundamentals, Kaggle or competition experience, LLM-related courses or projects, Git and data processing skills. 中文：加分项（非必须）：大模型安全评测（内容安全 / 越狱 / 偏见）、红队对抗（garak / PyRIT 等）、Python 自动化评测脚本、内容安全检测模型（LlamaGuard / ShieldLM 等）、AI 合规与认证经验、评测基准对齐（如 AILuminate）、语料库构建 / 数据标注、NLP / 机器学习基础、Kaggle / 竞赛经验、大模型相关课程 / 项目、Git / 数据处理能力。