SangHyuk Kim

Ph.D. Candidate & Research Engineer | Low-Level Systems + Medical AI

shk_pic.jpg

[Google Scholar] [GitHub] [LinkedIn]

I am a Ph.D. Candidate in Computer Science at the University of Massachusetts Boston, conducting research under Prof. Daniel Haehn in the Machine Psychology Group.

My research bridges the gap between theoretical Medical AI and production-grade Systems Engineering. I specialize in Volumetric Computer Vision (2.5D/3D) and Uncertainty Quantification, backed by over four years of industrial experience in low-level OS kernel development.

Prior to my doctoral studies, I worked as a Systems Kernel Engineer at TmaxOS. There, I architected custom graphics kernels and GDI+ interfaces to bridge Linux/Windows ABIs—a rigorous systems background that now underpins my work in building scalable, reliable healthcare AI.

🌟 Featured Research: Context-Aware 2.5D MRI Model

The Problem: Identifying the anatomical plane (Axial, Coronal, Sagittal) is standard for humans, but AI models fail on "near-skull" edge slices. These ambiguous images lack distinct features, leading to corrupted metadata and severe domain shift in large datasets.

Our Solution: We introduce a 2.5D Context-Aware Classifier. Instead of heavy 3D networks, we use a lightweight model that samples adjacent slices to learn local anatomical flow. This provides just enough context to resolve ambiguity, achieving >99% accuracy.

Clinical Impact: By gating this corrected metadata into a tumor detection pipeline, we reduced clinical misdiagnoses by 33.3%.

##

Research & Engineering Focus

  • Volumetric Medical Intelligence:

    • Context-Aware MRI Analysis: Architected the 2.5D CNN framework described above. This work reduced downstream brain tumor misdiagnoses by 33.3%.
    • Melanoma Detection: Developed uncertainty quantification pipelines that reduced diagnostic error rates by 40.5% (Accepted at IEEE ISBI 2025).

  • Full-Stack Research Deployment:

    • I build “Zero-Install” clinical tools. I engineer client-side inference pipelines using TensorFlow.js, React, and WebAssembly, enabling complex PyTorch models to run directly in the browser with no server latency.
    • Deployment: Real-Time MRI Plane Detector.

  • Systems & Kernel Optimization: * Optimizing high-performance computing tasks using C/C++ and OS internals (Linux/Win32) to ensure cross-platform operability for critical enterprise architecture.