AI算法专家

该职位来源于猎聘

• Design and own end-to-end enterprise RAG architectures for quality, engineering, specification, maintenance and corporate knowledge scenarios, ensuring accurate, traceable and scalable answers through fit-for-purpose technical design. 面向质量、工程、客户规格、设备维护和企业知识等场景，设计并负责端到端企业级RAG架构，通过适配业务的技术设计保证答案准确、可追溯并可规模化。
• Develop document ingestion and chunking strategies for unstructured and semi-structured content, including PDFs, tables, scanned documents, SOPs, specifications and reports, and determine chunk size, overlap, semantic boundaries, metadata and parent-child structures based on controlled experiments. 针对PDF、表格、扫描件、SOP、规格书和报告等非结构化及半结构化内容，开发文档导入与切片策略，并通过受控实验确定切片大小、重叠比例、语义边界、元数据以及父子切片结构。
• Optimize retrieval pipelines by selecting and benchmarking embedding models, vector databases, sparse and dense retrieval, hybrid search, query rewriting, filters and rerankers, in order to improve recall, precision and domain relevance. 通过选择和基准测试Embedding模型、向量数据库、稀疏与稠密检索、混合搜索、查询改写、过滤器及重排序模型，优化检索链路，提高召回率、精确率和领域相关性。
• Tune retrieval parameters such as Top-K, similarity thresholds, candidate pool size, reranking depth and context budget by analyzing evaluation datasets and failure cases rather than relying on fixed default values. 基于评估数据集和失败案例，对Top-K、相似度阈值、候选池大小、重排序深度和上下文预算等检索参数进行微调，而不是直接套用固定默认值。
• Build an end-to-end RAG evaluation framework covering retrieval recall, context precision, answer correctness, faithfulness, groundedness, citation accuracy, latency and operating cost, and use the results to drive systematic improvement. 建立端到端RAG评估体系，覆盖检索召回率、上下文精确率、答案正确性、忠实度、事实依据、引用准确性、延迟和运行成本，并用评估结果推动系统性改进。
• Engineer LLM and AI agent solutions with tool calling, workflow orchestration, memory and guardrails, and integrate them with approved enterprise data sources and systems while maintaining access control, security and auditability. 开发具备工具调用、工作流编排、记忆和安全护栏能力的大模型及AI智能体解决方案，并在确保访问控制、信息安全和可审计性的前提下，与获批的企业数据源和系统集成。
• Lead technical troubleshooting and root-cause analysis for hallucination, missing retrieval, irrelevant context, conflicting sources, long-context degradation, latency and cost issues, and implement measurable corrective actions. 针对幻觉、漏检、无关上下文、来源冲突、长上下文性能下降、延迟和成本等问题，主导技术诊断和根因分析，并实施可量化验证的改进措施。
• Define reusable engineering standards, reference architectures, code review criteria, test datasets and production readiness checks, while coaching AI Pilots and developers to deliver robust AI applications. 制定可复用的工程标准、参考架构、代码审查准则、测试数据集和生产上线检查要求，并指导AI Pilots及开发人员交付稳健的AI应用。
• Master's degree or above in Computer Science, Artificial Intelligence, Machine Learning, Natural Language Processing, Mathematics, Statistics, Software Engineering or a closely related discipline. A strong bachelor's degree combined with substantial relevant experience may be considered. 计算机科学、人工智能、机器学习、自然语言处理、数学、统计学、软件工程或紧密相关专业硕士及以上学历。具备扎实本科教育背景并拥有丰富相关经验者亦可考虑。
• Typically 5 or more years of relevant experience in AI, machine learning, NLP, search or data-intensive software engineering, including demonstrable hands-on experience delivering LLM, RAG or AI agent solutions beyond proof-of-concept stage. 通常要求5年及以上AI、机器学习、NLP、搜索或数据密集型软件工程相关经验，并具有可验证的实战经历，能够将LLM、RAG或AI智能体解决方案推进到超越概念验证的阶段。
• Experience in a leading technology company, AI company, cloud provider, research organization or large-scale digital product team is strongly preferred. 具有科技企业、AI公司、云服务商、研究机构或大型数字产品团队工作经验者优先。
• Experience with manufacturing, industrial engineering, quality management, semiconductor, electronics or enterprise knowledge systems is an advantage. 具有制造业、工业工程、质量管理、半导体、电子行业或企业知识系统经验者优先。