Deep Yunhai has secured a multi-million RMB angel round of financing, which was founded by the former Baidu search team | 36Kr Exclusive

Text by Wang Xinyi

36Kr has learned that recently, Deep Cloud Sea Intelligent Technology (Shanghai) Co., Ltd. (hereinafter referred to as "Deep Cloud Sea"), an AI marketing company focusing on GEO, has completed a multi-million-yuan angel round of financing, led by Huagong Tech (000988.SZ). The funds from this round will be mainly used for the iteration of self-developed technology base, overseas market expansion, product upgrading and the recruitment of AI talents.

Founded in June 2025, Deep Cloud Sea has a team of more than 70 people. Its core team comes from the former Baidu Search Department. Qian Cheng, the CEO, once served as the General Manager of Baidu Shanghai. He is also a serial entrepreneur. After leaving Baidu, he successively founded companies including Weizhi Interactive Media, Aiwai Golf, and Hi Xiu Auto, among which Aiwai Golf and Hi Xiu Auto were acquired by Ctrip and Didi successively. Yang Hailing, COO, has served as senior management such as COO of well-known enterprises including 3721, Baidu, HC360, DHgate, and World Union, with full-link commercial operation experience in leading Internet enterprises; Long Jie, CTO, once participated in the implementation of high-precision maps for Baidu's autonomous driving system.

According to the introduction, since its establishment for one year, Deep Cloud Sea has cumulatively served more than 3,000 enterprise customers, with an annual revenue of tens of millions of yuan.

An AI marketing tool that helps enterprises do business

The products of Deep Cloud Sea consist of two parts: the GEO RaaS system (including domestic version and overseas version) and the DeepWork desktop product.

AI enterprise intelligent marketing platform, Source: the enterprise

In simple terms, the GEO system optimizes through the generation engine, so that domestic and overseas AI tools such as ChatGPT, DeepSeek, and Doubao will give priority to recommending the client enterprises when outputting results.

Specifically, this system mainly includes the following contents:

• Diagnosis: Use Deep Cloud Sea's self-developed "Yunbao AI Brand Visibility Monitoring System" (AI visibility, brand sentiment, information source loopholes, competitor comparison) to evaluate the visibility of the brand in large models, and use the AI monitoring system to track the inclusion status, mentioned times and ranking of the brand in models such as Doubao, DeepSeek, and Yuanbao in real time;
• Optimization: Targeted optimize prompts according to the preferences of each platform, such as expanding intent words;
• Production: Layer the authoritative information sources, identify the remaining information gaps, and use the AI Lingque site building tool to transform the enterprise's official website into a form that is easier for large models to identify and crawl;
• Distribution: After the content goes through the dual compliance review by human and machine, it is distributed to various platform channels.

Another product, DeepWork, is an intelligent marketing workstation, positioned as "the customer's in-house AI marketing department". Different from all-in-one desktop Agents such as WorkBuddy, DeepWork is an Agent intelligent marketing workstation specially built for GEO. Centered on marketing work, this application is equipped with a diagnosis station, creation station, monitoring station, Skill Center and conversion station, covering the whole process from brand visibility diagnosis and tracking, publicity draft generation and distribution, to sales automation.

Qian Cheng believes that data security is a common concern for enterprises. Therefore, this application supports local deployment, offline availability and independent data control. Moreover, the platform will continuously accumulate personalized data, so that the system can understand the enterprise better as it is used more.

In terms of business model, Deep Cloud Sea takes the RaaS (Result as a Service) route, and it operates in both the domestic and overseas markets. In terms of industries, its customers are mainly distributed in machinery and equipment, chemical energy, electronics and electrical, education and other sectors.

The second half of GEO: The spotlight is on technology

"Large models are getting smarter, and their search traceability results are also expanding, from 10 entries to a maximum of 60 entries. At the same time, the ecosystem of traceability results is also changing. In addition to web pages, some AI applications such as Doubao can also find content such as Douyin videos as reference materials," said Qian Cheng. He drew a time dividing line for GEO: in the second half of this year, the industry has also entered a technical watershed, shifting from the traffic dividend period to the technology-driven period.

From the perspective of the product base, Deep Cloud Sea has independently developed the GEO marketing algorithm engine, and built an eight-layer enterprise knowledge graph covering entities, relationships, attributes, events, rules, FAQs, tags and training constraints. Qian Cheng summed up the technical route as "reverse engineering": observe what traceability checks the large model performs before outputting results, carry out reverse algorithm matching, and judge which sites are easier to be cited. Around this base, Deep Cloud Sea has applied for multiple software copyrights and patents.

GEO full-link operation system, Source: the enterprise

However, GEO will not completely replace SEO (Search Engine Optimization). Qian Cheng believes that GEO will become the standard configuration for enterprises to carry out AI marketing, and the industry is currently in the reshuffling phase.

Qian Cheng judges that the window period for start-ups is two to three years, and the reshuffling rules are clear: in the early stage of industry development, AI visibility can be achieved simply by publishing more articles that can be crawled by AI; but now the threshold for effect quantification has increased, and technology, algorithm and monitoring investment are necessary to ensure quantifiable delivery.

Qian Cheng told us that the existing large models are also changing their tendency to crawl information sources: the weight of self-media content being cited has dropped to 10%, while the weight of authoritative information sources continues to rise. Therefore, small GEO companies with substandard technology, insufficient investment and no operation capabilities may gradually fall behind.

This also makes Deep Cloud Sea constantly adjust its own route. After launching the DeepWork intelligent marketing workstation, they will also actively expand overseas markets. It is not enough to just be a GEO service provider. Qian Cheng hopes that Deep Cloud Sea can gradually become an AI global enterprise intelligent marketing platform in the future, allowing enterprises to improve their visibility in the AI ecosystem, while enabling intelligent "assistants" to complete other marketing chains, so that AI can help enterprises grow their business continuously.