cvt engineering offers AI engineering consulting for R&D teams that work with GeoDict, the software for digital material R&D by Math2Market, or with other simulation software. The work covers AI agents with which simulations are set up and run, macros are written and result files are evaluated into reports. With the GeoDict Copilot, cvt engineering has built such an agent for GeoDict itself.

Scope of the work
Together with your simulation team I build what an agent needs to work with your tools:
- Your simulation workflows written as agent skills, with the procedure and the checks your team applies.
- Your existing GeoPy (GeoPython) macros and GeoApps made callable by an agent.
- Your in-house agent system connected to GeoDict through an MCP server or an API.
- Module knowledge, material data and company conventions collected in a knowledge base for the agent.
- Tests with fixed tasks and known results, so that a change to skills or model can be checked.
- Models compared on your tasks: commercial, hosted in Europe or on your own servers.
Agents for other simulation software are built the same way when the software offers a scripting interface or an API. The agents operate the simulation software and its solver; surrogate models that predict results without the solver are a different field and not part of this offer.
Results for the company
Skills, knowledge base and tests are text files that stay with the company. The macros are GeoPy files in the project folder and run in GeoDict without the agent.
Background
I worked as an application engineer for GeoDict at Math2Market and have run cvt engineering since 2018. For eight years I have been developing my own tools for this work, and the GeoDict Copilot builds on them.
First conversation
A first conversation starts from one workflow your team runs often. A short description of it is enough.