Long benchmarked against Palantir, Ren Xinqi, a Peking University alumnus, has secured another round of funding.

On September 11, it is learned from ChinaVenture that Yue Dian Technology, an enterprise-level AI company, has completed a Pre-A round of financing of tens of millions of RMB, led by Infore Tec and followed by Tsinghua Venture Capital. The funds from this round will be mainly used for AIGO model training, promoting the continuous evolution of AI FDE Model, improving the RSI closed loop, and accelerating the market replication of verified scenarios.

While most enterprise-level AI companies are still worried about POC conversion and subscription renewal, this four-year-old company that has achieved break-even has delivered a distinctive performance report: it has nearly 100 paying customers in total, with a compound annual growth rate of more than 300% for both orders and revenue from 2023 to 2025, and its revenue in the first half of 2026 has exceeded the total revenue of 2025.

Spin off from the knowledge graph product line of Minglue Group, Yue Dian Technology has long taken Palantir as its product reference. It focuses on enterprise-level decision intelligence, targeting private deployment environments of enterprises, enabling AI to participate in data analysis, business judgment and task execution. In short, they want AI to connect the entire chain of cross-system data query, judgment and execution, to take the place of the "senior veteran who knows everything" in enterprises to finish the work.

"We are not moving Palantir's business territory to China," Ren Xinqi, Founder and CEO of Yue Dian Technology, said to ChinaVenture. The company retains Palantir's product idea of using private data to support complex analysis, decision-making and execution, and is committed to solving the unique problems of Chinese enterprises: enterprise AI asset iteration, private AI system construction, and core business scenario adaptation.

Founded by Peking University Alumni, Benchmarked against Palantir

The founding of Yue Dian Technology is inseparable from Ren Xinqi's long-term research on knowledge engineering, knowledge graph and enterprise data products, and also stems from a problem that has troubled him for a long time.

Back in 2020, when he reviewed the projects in hand, he found a phenomenon: the same knowledge graph and analysis platform showed significantly different effects in different customers. In Ren Xinqi's view, the variable lies in whether there are experts who truly understand the business inside the customer. "Experienced people can use the system to find clues, establish connections and make judgments. If the user lacks business experience, the same platform can hardly exert the same value."

Behind this judgment is Ren Xinqi's nearly 20 years of professional accumulation. He studied in the Department of Computer Science of Peking University for both his bachelor's and master's degrees. He started to study knowledge engineering as early as in the Peking University Artificial Intelligence Laboratory, and got access to Ontology, knowledge representation and reasoning. After that, he was engaged in the practical implementation of search engines, Knowledge Graph and large-scale data at Baidu.

In 2014, he participated in the founding of Minglue Data as a technical partner, and incubated and expanded the enterprise knowledge graph product SCOPA from scratch. According to Ren Xinqi's recollection, SCOPA took Palantir as an important product reference from the very beginning of its establishment, but the team did not copy all its businesses. Instead, it chose private data, in-depth analysis and business decision-making that are more suitable for its own capabilities and the Chinese market.

The SCOPA enterprise knowledge graph platform organizes the originally scattered data, establishes entities and relationships, and then brings these capabilities into complex businesses such as finance. The capability portfolio composed of knowledge engineering, enterprise data engineering, AI technology, ToB products and large-scale enterprise project delivery has also become the confidence for Yue Dian Technology to continue its subsequent entrepreneurship.

In order to truly complete the closed loop of analysis, decision-making and action, Ren Xinqi founded Yue Dian Technology in Beijing in June 2022 and developed it independently. In addition to considering that businesses involving private data of large and medium-sized enterprises are more suitable for long-term development in China, he believes that to solve the "expert dependence" problem, it is necessary to use AI to reconstruct the product and technical architecture. "This kind of re-invention exploration is more suitable to be carried out in a new company."

After becoming independent, Yue Dian Technology made two choices. Technically, instead of blindly continuously maintaining a set of traditional knowledge graph products, it connected AI to the original data and knowledge system. On the market, they no longer only focus on the familiar security and finance fields, but gradually shift their focus to general industrial scenarios such as manufacturing, energy and transportation.

Their direction has gradually become clear, and they continue to take Palantir as a product reference to learn from its method of using private data to support in-depth analysis, decision-making and action. The company plans to further promote the data organization problems solved by knowledge graphs in the past to business execution in the industrial and private deployment scenarios of Chinese enterprises.

Ontology has also been placed back in the core position of the product, undertaking the task of organizing business objects such as customers, orders, products and equipment, as well as the relationships, rules and executable actions between them. "Large language models know how the world works, but they don't know how this enterprise works. Ontology is to tell the latter to it." Ren Xinqi said frankly that he hopes to reduce the uncertainty of large language models in the core processes of enterprises, so that people and AI can work based on the same set of business structures.

The emergence of large language models has provided new technical conditions for this line. In 2023, Yue Dian Technology launched the R&D of domain-specific large models. The next year, the team began to try to let AI undertake part of the data processing, knowledge extraction and scenario construction work that previously required engineers to complete through the AI FDE method, and the product was later upgraded to Knora-AI.

As the product direction gradually became clear, they completed the angel round of financing exclusively invested by Qiming Venture Partners in December 2024. Qiming Venture Partners believes that with the evolution of AI large model capabilities and the further expansion of application boundaries, the market has strong demand for end-to-end capabilities, and Yue Dian Technology has significant advantages in the combination of AI and data intelligence, with a complete end-to-end product and great potential in data intelligence and large model applications.

By enabling large language models to undertake more standardized work, Yue Dian Technology further integrated Ontology, Agent and autonomous execution capabilities into the same product system in 2026. Infore Tec and Tsinghua Venture Capital also followed the long-term accumulated technical route of this company and finally became its shareholders.

"The key to enterprise AI construction is to integrate private knowledge to form core assets oriented to AI, people and organizations, in which Ontology plays an irreplaceable role." In the view of Wang Shu, an investor from Infore Tec, "The team of Yue Dian Technology is the first team in China to propose enterprise-level knowledge graph, carry out long-term practice of Ontology and implement it on a large scale."

Enable AI to Make Judgments and Take Actions Autonomously, the Company Has Achieved Break-even

In addition to making enterprise software less dependent on experts outside the system, Yue Dian Technology is also enabling AI to freely call data, knowledge and business capabilities.

What will happen when a manufacturing enterprise has quality problems? Customer information is in the CRM, production batches are in the MES, raw materials and suppliers are in the ERP, and processes and designs are in the PLM. In the past, this required people in different positions to query separately, and then business personnel combined their experience to form a judgment.

Yue Dian Technology does not just want AI to answer questions. Instead, it first organizes these business information through Ontology, and then lets Agent complete query, task disassembly, tool invocation and execution, connecting the processes that previously required multiple people to work in relay.

The effect is directly reflected in time and cost. Ren Xinqi gave an example that in a set of quality traceability process, it originally required multiple people to spend several days for cross-system query. After accessing Ontology and multiple Agents, one employee can drive the entire process, and the query time is shortened to 5 minutes.

Based on Ontology and enterprise data capabilities, Yue Dian Technology will further transform business goals into tasks that Agents can execute through task disassembly, reasoning and action arrangement. In addition to deepening specific scenarios such as quality management, equipment operation and maintenance, and supply chain, they will also ensure that data does not leave the domain and the process is traceable, which is an unavoidable hard threshold for state-owned enterprises and leading manufacturing enterprises.

This also enables enterprises to obtain two types of value: one is the measurable explicit value, such as reducing manual input, improving efficiency, reducing errors, shortening business cycles, etc.; the other is implicit value, including reducing cross-departmental collaboration costs, precipitating the experience of senior engineers and business experts, and reducing knowledge loss caused by personnel turnover.

"The real vitality of enterprise AI does not lie in whether it can get several large customers and form several large orders, but in whether it can enter the core business of the enterprise, maintain stickiness in high-value scenarios, and continuously solve complex problems." According to Ren Xinqi's disclosure, in the quality management scenario of a leading panel manufacturer, Knora-AI helped the customer increase the yield by 3‰, with the corresponding revenue exceeding 20 million yuan.

At present, Yue Dian Technology has nearly 100 paying customers in total, most of which are large enterprises. Leading private manufacturing enterprises mostly adopt software authorization and subscription methods, while state-owned enterprises usually choose a combination of software, project implementation and services, and part of the implementation work can also be completed by integrators or third parties.

"From 2023 to 2025, the compound annual growth rate of both orders and revenue of Yue Dian Technology exceeded 300%. The revenue in the first half of 2026 has exceeded the total revenue of 2025." Ren Xinqi said frankly that the company has achieved break-even, and it is expected that the revenue in 2026 will exceed 100 million yuan and continue to make profits.

In the coming year, they hope to further combine products with large models, improve the self-iteration capability in real scenarios such as production manufacturing and product R&D, gradually reduce manual intervention, and realize the RSI (Recursive Self-Iteration) closed loop. "The short-term goal is to form a closed loop that does not require continuous manual participation in some production environments."

More than ten years ago, Ren Xinqi and his team worked on enterprise knowledge graphs to organize scattered data and relationships for business experts. Today, Yue Dian Technology takes a further step, putting the capabilities originally scattered in data, processes, rules and expert experience into the system, enabling AI to participate in judgment and action, and forming an enterprise-level product that is easier to replicate.

This article is from the WeChat Official Account "ChinaVenture", author: Lu Zhigao, authorized for release by 36Kr.