cuOpt — математическая постановка задачи оптимизации

★ 7.6 · content

cuopt-numerical-optimization-formulation is a Claude Code skill that translates a plain-text problem description into a rigorous mathematical formulation for LP, MILP, and QP — without writing any API code. It classifies every sentence in the problem statement into four categories: parameters, decisions, constraints, and objective, and explicitly surfaces implicit constraints hidden in committed phrasing ("plans to produce") or implicit objectives ("determine the plan" when costs are given). The skill covers LP (continuous variables), MILP (integer and binary variables for scheduling and facility selection), and QP (quadratic objectives for portfolio variance or least squares). Developed by the NVIDIA cuOpt Team (Apache-2.0, version 26.10.00), it is intended as a formulation step before moving to any solver API or implementation code.