Can Bogoclu
teacher of machines. applier of science. designer of things.
I build models of systems that have physics in them, and I care about knowing when those models are wrong.
My PhD was on multi-objective design optimization under uncertainty: how to design a physical system so it meets a reliability target when every input is noisy. I applied it to electric motor design with Bosch, reaching an expected reliability of one defect per million, and later led ML projects for DLR, ZF and others spanning surrogate modelling, reinforcement learning and computer vision. These days I work on machine learning for logistics at scale, where the problem underneath is the same: sample efficiency, uncertainty quantification, and models whose errors have physical consequences.
Software
- experiment-design — design of experiments and constrained space-filling sampling
- uncertainty-propagation — uncertainty propagation and reliability analysis
- pirl — model-based reinforcement learning with probabilistic dynamics models
Besides research and mentoring, I like
- Learning new things
- Tackling problems without an apparent solution
- Good visualizations
- Curious people
- Beer