MoenoIto

I am a first-year master's student in the Design Thinking & Science program at the University of Electro-Communications (Shono Lab), specializing in deep learning for distribution-shift robustness. My core research tackles domain shift in CT-based lung disease classification — specifically, robustness to variation in windowing parameters (WW/WL) across hospitals, which silently degrades classifier performance in real clinical deployment. I have compared statistical harmonization, CycleGAN-based translation, and hybrid preprocessing pipelines, and designed windowing-augmentation strategies to close the cross-domain gap. I approach robustness as a general problem rather than a medical-imaging-specific one: my aim is to build models that behave predictably under distribution shift in any industrial setting. I am currently interning at SmartAvatar B.V. (Delft, Netherlands), working on a trust-scoring system for nodes in a distributed AI inference network. I care about the gap between benchmark numbers and real-world reliability, and I want to work where that gap has consequences.