Together with the NETZSCH Group, we designed LabV, took a key role in its development, and brought it to market as a product that helps laboratories make better use of their data:

• Data aggregation: LabV brings together data from laboratory instruments on a central data management platform. A smart mapper is used to process the different data formats and store them in a homogenized format.
• Data integration: The merged data is organized in projects and can be combined with other data sources. This allows project details to be obtained from ERP systems or links to be created between the laboratory data and the associated MES data.
• Data analysis: Data can be searched for specifically within the platform and further analysis can be carried out, e.g., by adding statistical evaluations or graphical representation.
• Data preparation: Project results can be automatically exported to test reports or comparable forms of documentation using Office templates.
• LLM support: With the help of the LabV assistant, further complex searches and analyses can be carried out in the data using natural language instructions. LabV integrates the database with the analysis functions of Wolfram Mathematica and makes it possible to use documents for further research.

Within the project, we navigated each phase of the complete development cycle of a professional software product, ensuring a systematic and organized approach:

‍

• Market analysis and business model development: Before software development began, the market was thoroughly analyzed, user needs were identified, and a requirements profile was created which provided a basis for the business model.
• Minimum Viable Product (MVP): An MVP was then created as a functional demonstrator, which was validated both at trade shows and through initial customer use.
• Market-ready product: Based on feedback, the product was refined in terms of both functionality and user experience to ensure professional marketability.
• Ongoing development: Continuous product enhancement is professionally managed, focusing on adding new features and maintaining the technological foundation to ensure easy scalability and maintainability.

In this project, a scouting process was conducted to compile a long list of over 50 relevant AI use cases across four different industries, each briefly described according to defined criteria.

‍

A particular focus was on data strategies used in other industries, how data is interconnected, and how companies collaborate with upstream and downstream partners along the value chain to create cross-company platforms.

‍

Six use cases selected by the client were explored in greater depth through deep dives and qualitative interviews.

When RFIs / RFQs need to be filled out, it is a very time-consuming back and forth between different people such as subject matter experts, account managers, and the client themselves. We have developed an LLM-based solution for a medical solutions provider to pre-populate RFIs and RFQs. The entire workflow was streamlined to best integrate the AI application into the process and maximize value.

‍

Key elements of the project:

• Collection of a dataset of previously asked questions and their corresponding answers.
• Processing of the data to create an index that allows for searching similar questions and answers.
• Integration of new questions from RFIs/RFQs with context retrieved from the index.
• Utilization of a pretrained LLM to answer the new questions based on the retrieved context.

As a result, new RFI/RFQ forms can be processed much faster with the help of the LLM-supported app, as the questionnaires can be pre-filled and only require a final review.

‍

Together with the Hong Kong Industrial Artificial Intelligence & Robotics Centre (FLAIR), we developed the following AI solutions in cooperation with various industry partners:

• Chat With Your Dashboard: LLM-based chat application that can answer questions about data visualized on a dashboard. For example: What was the average OEE in the last week? Was there an anomaly for temperature zone 1 in the last hour?
• AI-Assisted Derivation of Production Parameters: Web app that allows training of selected models on data from an injection molding process.  Subsequently, the trained models can be employed to predict product quality KPIs from the settings of the injection molding machine.
• Maintenance Guidance: LLM-powered application, which has a general understanding of wear components, their failure mode, their causes and the corresponding maintenance and servicing measures to guide and support maintenance personnel.  
• AI Assistant for Assembly Automation Engineers: LLM-based chat application that helps automation engineers make the right machine and equipment selection when designing a new assembly line.