内容来源：https://aibusiness.com/foundation-models/startup-large-tabular-model-spreadsheet-data

内容总结：

近日，旧金山初创公司Fundamental宣布结束隐匿模式，并成功获得2.55亿美元融资，其中包括3000万美元的种子轮融资和由康涅狄格州风投公司Oak HC/FT领投的2.25亿美元A轮融资。该公司专注于解决生成式AI在结构化数据处理领域的短板，推出了名为Nexus的大型表格模型。

Fundamental指出，当前主流大语言模型主要擅长处理文本、图像等非结构化序列数据，但在解读企业日常使用的电子表格等非线性、非序列化表格数据时存在局限。Nexus模型则专门针对这一痛点设计，其核心团队来自谷歌DeepMind，基于数十亿表格数据集在亚马逊SageMaker HyperPod上训练而成，能够自动解析表格中行列间的复杂关系与交互模式。

公司联合创始人兼CEO杰里米·弗兰克尔表示，Nexus是“专为挖掘全球最具价值数据——支撑各行业企业决策的数十亿表格——而构建的通用基础模型”，并将其定位为“商业决策的操作系统”。该工具可无缝接入客户现有数据栈，无需人工训练即可自动学习数据结构与规律。

投资方Oak HC/FT管理合伙人安妮·拉蒙特评价称：“Fundamental模型的意义重大——结构化关系型数据至今尚未充分受益于深度学习革命。该技术能够预测从金融欺诈、患者再住院率到能源价格等多元场景，几乎能为所有行业提供支持。”

目前，Fundamental已与亚马逊云科技达成合作，客户可在AWS环境中直接部署Nexus模型。此次融资除领投方外，还吸引了Valor Equity Partners、Battery Ventures、Salesforce Ventures等多家风投机构及Perplexity AI首席执行官阿拉温德·斯里尼瓦等天使投资人参与。

中文翻译：

由谷歌云赞助
选择您的首个生成式AI应用场景
要开始应用生成式AI，首先应关注能够优化人类信息交互体验的领域。
该公司的AI工具专为解析结构化数据而设计，这正是大语言模型难以处理的领域。
总部位于旧金山的初创企业Fundamental开发了一种从企业数据中驱动预测的技术，近期以2.55亿美元融资结束隐匿运营状态。该公司通过3000万美元种子轮融资和2.25亿美元A轮融资（由康涅狄格州的Oak HC/FT领投）获得资金支持，参投方包括Valor Equity Partners、Battery Ventures、Salesforce Ventures和Hetz Ventures等风投机构，以及Perplexity AI首席执行官Aravind Srinivas在内的多位天使投资人。

Fundamental的核心产品是一款能解析电子表格的AI工具，这正是大语言模型的薄弱环节。该公司将这项创新称为"大型表格模型Nexus"。联合创始人兼首席执行官杰里米·弗兰克尔在新闻稿中表示："我们构建的通用基础模型专门用于挖掘全球最具价值的数据——支撑各行业企业预测的数十亿张数据表。Nexus是商业决策的操作系统。"

据该公司介绍，大语言模型及相关架构通常针对文本、图像、视频等非结构化序列数据优化，而难以适配由非线性、非序列关系构成的表格数据——这类数据正是大型企业常用的电子表格核心。Nexus旨在解决这一痛点。该工具由谷歌DeepMind前成员研发，基于"数十亿表格数据集"在亚马逊SageMaker HyperPod上从头训练，使其能解析行列间的非线性关系与交互作用。

该公司还宣称可轻松集成至客户数据栈，连接后无需人工训练即可自动学习相关数据的模式与结构。投资方Oak HC/FT管理合伙人安妮·拉蒙特在新闻稿中评价："Fundamental模型的意义怎么强调都不为过——结构化关系型数据尚未充分受益于深度学习革命。其预测能力涵盖金融欺诈、患者再入院率、能源价格等多元场景，使该公司几乎能为所有行业提供支持。"

Fundamental已与亚马逊云科技建立合作，客户可在AWS环境中部署Nexus系统。

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
The company's AI tool was built to make sense of structured data, an area where large language models struggle.
Fundamental, a San-Francisco-based startup that has developed a way to drive predictions from enterprise data, emerged from stealth recently with $255 million in funding.
The company has raised a $30 million seed round plus a $225 Series A led by Connecticut's Oak HC/FT, with participation from other venture capital outfits including Valor Equity Partners, Battery Ventures, Salesforce Ventures and Hetz Ventures, plus a handful of angel investors such as Perplexity AI’s CEO Aravind Srinivas.
At the heart of the Fundamental proposition is an AI tool that can make sense of spreadsheets, an area where large language models struggle. The company is billing its innovation as a large tabular model called Nexus.
"We've built a generalized foundation model specifically to leverage the world's most valuable data: the billions of tables that underpin predictions in every enterprise, across every vertical," Jeremy Fraenkel, CEO and co-founder, said in a press release. "Nexus is the OS for business decisions."
LLMs and related architectures are generally optimized for unstructured, sequential data such as text, images and video, according to the company. They're less suited to tabular data, which is made up of non-sequential, non-linear relationships including data found in spreadsheets and tables commonly used at large enterprises.
Nexus aims to address this pain point. The tool, which Fundamental said has been built by alumni of Google's DeepMind, was developed from the ground up on "billions of tabular datasets" and trained on Amazon SageMaker HyperPod, endowing it with the ability to make sense of the non-linear relationships and interactions that exist across rows and columns.
The company also claims to easily integrate into customer data stacks. Once connected, the tool automatically starts learning patterns and structures of the relevant data with no manual training.
"The significance of Fundamental's model is hard to overstate — structured, relational data has yet to see the benefits of the deep learning revolution," said Annie Lamont, managing partner at investor Oak HC/FT, in the press release. "[The] ability to predict anything from financial fraud to hospital readmission to energy prices positions the company to support virtually every industry and sector."
Fundamental has also forged a partnership with Amazon Web Services that will allow customers to deploy Nexus in their AWS environment.