The problem
Quantum kernel methods depend on the circuit used to encode classical data. Choosing that circuit by hand restricts the search to architectures anticipated by the designer.
The approach
An evolutionary algorithm searches over circuit structures built from parameterised rotations, entangling gates and reusable motifs. Candidate kernels are evaluated through classification experiments, with attention to generalisation and computational cost.
Research direction
The work investigates fitness selection, sparse kernel evaluation and circuit structures that transfer across qubit counts. Research is ongoing; final results and reproducible experiments will be added as the work develops.