Supervisor: Jeanette Hellgren Kotaleski (EECS, KTH). I am registered at CBH at KTH and Dept. of Neuroscience at KI in the KI-KTH-joint PhD program. This application is for a separate PhD-student project allocation, intended to support method development and production runs for my thesis work without drawing on shared group allocations. Fitting parameters to detailed multicompartment models is notoriously difficult, typically requiring high-performance computing resources to explore their vast parameter space. Biological neurons, however, control this parameter space to maintain their function across diverse perturbations. Using this phenomenon as inspiration, I have developed an optimization framework, HOB, where model parameters are obtained not only based on fitness metrics, but also the corresponding internal state of the model. The expected outcome is a set of optimized biophysical neuron models and GPU-accelerated simulation workflows for investigating how cellular mechanisms shape network activity in basal ganglia circuits. In particular, since HOB is based on JAX, it'd be helpful to test batched optimizations on GPUs.