What stands out most to me is the shift from slow, batch-based numerical simulation to real-time physics feedback. By training surrogate models directly on 3D geometric meshes, it reduces simulation turnaround from hours or days to milliseconds. That speed-up lets engineers explore thousands of design variants early in the development cycle, instead of being held back by solver bottlenecks. Review collected by and hosted on G2.com.

The main challenge is the platform’s dependence on data. It can deliver tremendous value once it’s properly trained, but reaching that stage requires a large volume of high-quality, standardized simulation data. For teams dealing with fragmented legacy datasets—or working on brand-new design architectures with little to no simulation history—the upfront effort to generate training data and put solid MLOps workflows in place can be significant. Review collected by and hosted on G2.com.