Every step of a founder's journey deserves a record.Founders · AI · PolicyCHINAOPC / ENGLISH EDITION

NovasNexus Manufacturing FDE: Bringing Knowledge Bases into Employee Training

According to an interview with Datawhale, NovasNexus co-founder Cai Jieru carried out AI training process projects for German manufacturing companies. This article starts from the moment when newcomers raise questions, and discusses how to accept knowledge application conditions, course progress, and teacher feedback together. The case is self-reported by the interviewee. The customer is not a small OPC team, and the test ratio is not a common standard for other companies.

Chinese edition published 23 Oct 2026 · 阅读中文原文 ↗

Source timingThe Datawhale case collection is due on September 6, 2026; read the text and image title page of NO.01 in full, and check until October 23, 2026.

The project object is new employee training, not just document query

Datawhale's "FDE Cases 100" is the first to include an interview with Cai Jieru, and the title page indicates that she is the co-founder of NovasNexus. According to interviewees, the project collates experience and knowledge for German manufacturing companies and integrates AI capabilities into existing training platforms. The following research is based on this self-report and does not mean that ChinaOPC has entered the customer's on-site verification.

This article chooses this case to discuss how the service team completes enterprise delivery. The customer has multiple production lines and cannot be classified as an OPC company; the complete employment and project finance of NovasNexus have not been confirmed in this study.

The position where the newcomer is stuck determines how the information is organized.

For teams that are preparing to provide similar services, you can first ask an observation question: at what time do new people need help? If the same abbreviation represents different objects in different production lines, only the entry explanation may not be enough; if the problem comes from not knowing whether the information is still applicable, continuing to increase the number of documents may not be useful.

ChinaOPC recommends logging the issue along with the product, equipment, data version, and identifiable personnel. This is a survey form design to be verified, not a restoration of the case system fields. The purpose is to make the answers generated later have applicable conditions, rather than compressing all the experience of old masters into standard answers without context.

A course question and answer must have a path back to human beings

Interviewees explained that AI capabilities are connected with course learning, progress viewing, and Q&A. From this, further research can be conducted: after newbies get an explanation, do they know which information to look for next, who takes over when they still have questions, and whether the instructor can see recurring problems.

For example, the team can work with the customer to arrange processing statuses such as "resolved", "needs supplementary information", "needs guidance and confirmation", etc. When it comes to actual equipment operation and safety matters, explanatory answers should not be taken as authorization for self-operation. The company's existing operating procedures and job responsibilities need to continue to have a clear position in the system.

It is not advisable to check whether the answer is satisfactory or not during the demonstration.

Technical inspection and business inspection can be designed separately. The technical side looks at whether the references come from applicable materials and whether key qualifications are missing; the business side looks at whether newcomers can continue to learn and whether the master can discover unmastered content. Both should use tasks and samples agreed with the client.

The parsing and recall testing mentioned in the original interview were the practices under project conditions at that time. This article does not translate the corresponding proportions into industry passing standards, nor does it replace correct judgment with language fluency. Actual goals need to be determined based on the consequences of errors, available information, and human inspection capabilities.

Count ongoing maintenance into service scope

Changes in teaching materials, equipment changes, and new experiences added by teachers will continue to affect the answers. The service team should agree with the customer on information updates, issue review, and exit handover arrangements. For light teams, whether they can undertake this part of long-term work is an important condition for deciding whether to undertake the project. Quotations cannot be based solely on the workload of initially building the page.

Extended reading: Acceptance and handoff of FDE delivery; Evidence chain of project records.

Sources and editorial notes

AI-assisted organizing and writing. Directly read the first interview with Datawhale, independently organized and analyzed rather than rewritten paragraph by paragraph based on questions and answers; the source statement has not been independently audited. The record fields, status design and security boundary discussions in this article are ChinaOPC recommendations and do not pretend to be existing system functions. English translation of the Chinese edition, prepared with machine assistance and editorial checks. Source verification dates refer to the original Chinese research.

Source titles and publisher names are retained in their original language. Refer to the linked originals for official documents and current requirements.

← Back to FDE CasesEditorial standards ↗