invrs.io aims to advance AI-guided design, focusing initially on optics---a space critical for components in AR/VR, datacenters, autonomous vehicles, and beyond. We are developing an ecosystem that is accessible to everyone from AI scientists to optimization researchers and optics designers. The main parts of the ecosystem are,
- invrs-gym: A collection of optics design challenges with a common API, intended to facilitate research and development of new design methods. For optics designers, the gym also serves as an example of how to structure design challenges, so that they can be used with other parts of the ecosystem.
- leaderboard: A database of solutions to gym challenges. Submissions are welcome!
- invrs-opt: Optimization algorithms with a common API, which currently include well-known schemes such as L-BFGS-B.
- totypes: Defines custom types used in the gym, and generally applicable to AI-guided design, topology optimization, and inverse design.
- invrs-utils: Utilities, including those which simplify the running and analysis of experiments.
We are just getting started, so expect things to change. If you are interested to collaborate please contact @mfschubert.