NVIDIA Open-Sources Auto Ontology for Enterprise Data Semantics
The shift is from tribal analyst knowledge to certified infrastructure: once business definitions are codified, sales, finance and operations agents share one auditable source of truth instead of each inventing its own metrics. Teams building text-to-SQL or data agents should evaluate it against hand-written dbt metric layers, since the Certify and Validate steps directly target hallucinated numbers. NVIDIA released it open-source, so enterprises can self-host the semantic layer alongside their warehouses.
NVIDIA has open-sourced Auto Ontology, a semantic layer that teaches AI models the business meaning of enterprise data. It addresses cases like querying Q3 East China revenue, where no such field exists and the model must know which columns compute revenue, which regions count as East China, and which tables can be joined. The workflow has four steps: Discover reads database metadata, Ground maps terms like customer and revenue to tables and columns, Certify lets business staff verify definitions, and Validate refuses queries the data cannot support, such as ad ROI without ad-spend data. The result is a certified, auditable semantic layer that agents can query instead of guessing.
英伟达开源的一个能理解企业数据的AI语义大脑：Auto Ontology，让AI真正理解数据的业务含义，从而给出可信可审计的回复
它解决几个问题，大模型会查表但不懂业务口径
比如说，“上季度华东收入多少”，数据库里没有一个叫“华东收入”的字段
模型需要知道收入是由哪些字段计算、华东对应哪些地区、时间字段应该用下单日发货日还是开票日、哪些表能连哪些不能连
Auto-Ontology就是先把这些定义整理出来，让模型按已认证口径查询
还一个，数据分析师在重复解释上浪费时间
“这个收入为什么不等于财务报表里的收入”、“这张表应该和哪张表连”等等
Auto-Ontology就可以把这些重复知识固化成可查询、可审核、可复用的语义层，减少每个AI、每个用户都重新猜一遍的情况
核心工作流包括这几步： 1、Discover自动发现数据结构 读取数据库元数据，识别业务实体、字段、关系以及潜在的语义关联
2、Ground语义落地 把“客户”“订单”“收入”等业务术语映射到具体表和列
3、Certify人工认证 业务人员可以检查、修正业务术语、SQL指标以及表关系，对经过核验的语义定义进行认证和管理
4、Validate校验查询 查询前先判断当前数据是否真的支持这个问题，比如计算广告回报率时，如果缺少广告支出数据，它会要求澄清，不是说硬算一个广告回报率
最大的价值应该是它把“企业数据是什么意思”从分析师脑子里的隐性经验，变成了AI可调用、可认证、可审计的基础设施
你的销售Agent、财务Agent、运营Agent也不必各自盲猜含义，可以基于统一的业务定义理解查询数据
#AutoOntology #企业数据AI #企业AI
