[QST] Plan to support Convex Optimization?

I was recently hanging out with some NVIDI folks at SciPy, and we were discussing the wide range of scientific computing tools that are now supported for GPU acceleration. I didn't see any optimization libraries so i asked around and was directed to CuOpt. The libraries look awesome but it seems to be bound to linear programming (LP), including integer (ILP) and mixed-integer (MILP), as well as combinatorial problems like the Vehicle routing problem (VRP). For the foreseeable future, is the focus on supporting traditional and emerging logistics problems, or is there a plan to support methods in non-linear optimization? i am thinking the equivalent of a tool for GPU accelerated disciplined convex programming (DCP) potentially with APIs matched to CVXPY. I'd also totally understand if that would be a different project altogether; maybe you know of one? No, urgency -- but I am told you folks are the best ones to ask. Thanks in advance.

Start by reviewing CuOpt’s current support for LP, ILP, MILP, and VRP, then compare the requested non-linear optimization and CVXPY-compatible DCP scope. Done would be a documented decision about whether this belongs in CuOpt or another project; the issue names no files or tests.