What drove the decision to develop an AI-based use case?

HQS develops software for the simulation of molecules and materials. The description of the systems to be simulated requires a large number of parameters, whose names and conventions are often inconsistent from one field to another. In addition, there are interactions and interdependencies between the various parameters, so that even experts require intensive training to use the tools. The manual input of the numerous parameters remains a time-consuming and error-prone task. With the Modeling Assistant, we want to use generative AI to make our software tools more intuitive and easier to use. With just a few keystrokes from the user, the Modeling Assistant can set up a simulation, validate the input data for the simulation software, and start the simulation. This makes it much easier to get started and work with our tools.

 

What were your expectations or requirements for the AI application?

A key requirement is the integration of domain knowledge. On the one hand, this includes knowledge about our software tools, in particular specialized knowledge about the conventions and dependencies of the parameters. On the other hand, the user should be able to upload specific knowledge so that the AI-assistant can incorporate it. 

The other requirements are related to the integration into the user's workflow. Our Python-based software tools are used by our customers in programming environments. So, the assistant needs to be able to invoke the tools in that environment. Moreover, it is helpful if the assistant can directly create Python objects and make them available to the user. The integration of visual input and the execution of mathematical operations further enables the assistant to set up simulations based on formulas from scientific publications. 

 

What frameworks/methods did you rely on for development?

Our application is based on the OpenAI API and uses, for example, the "vision" and "function calling" functionalities. It is important to us that the code of our software tools is not send to OpenAI, but only the description and the requirements of the input values for the various functionalities. 

However, we are also working on combining open-source AI models to provide alternatives to OpenAI. 

 

The use of AI solutions must also be accompanied by a change in competences. How have you dealt with this challenge in respect of your employees?

Most of our team members are experts in a scientific discipline or in software development. The initiative to develop AI tools came from several employees with an interested in this topic. The skills required to develop HQS's other products overlap strongly with those required to develop the HQS Modeling Assistant. In addition, we have trained all areas of the company in the use of AI tools through internal training. 

 

In what timeframe did they implement the AI solution in your company/your client‘s company?

Initial ideas were discussed and tested shortly after the first free trial version of ChatGPT was released. The specific development of the HQS Modeling Assistant has been going on for about nine months now. However, the improvement and development continues. 

A conclusion to your AI solution:

The use of innovative AI tools is a real asset in the development of our simulation software. It allows us to remove a major hurdle in the use of such software. 

 

One final piece of advice/tip to other entrepreneurs looking to apply AI:

AI offers endless possibilities. But not all of them are useful in practice.  

My advice is to set clear goals after a short exploration phase and to constantly check and question the added value of the solutions during development. 

 

 

https://quantumsimulations.de/hqs-modeling-assistant

 

Picture: HQS