I work on Computer Vision systems under different computational constraints,
ranging from large-scale medical imaging trained on HPC clusters to deployable Edge AI on embedded devices.

My interests lie in system-level AI, where data pipelines, model architectures, training infrastructure, and hardware deployment are co-designed as a unified system.

• 🎓 Electrical and Computer Engineering @ The Ohio State University
• 📈 GPA: 3.78
• 🔬 Experience in research projects, internships, and team competitions
• 🧠 Strong background in Computer Vision, Medical Image Computing, and Deep Learning
• 🚀 Hands-on experience with HPC-based multi-node, multi-GPU training
• ⚙️ Practical deployment experience on embedded and resource-constrained devices
• 🎨 Front-end development experience with Figma, UI design, and Git-based collaboration

• Medical Image Computing & Computational Pathology
• Computer Vision (CNNs & Vision Transformers)
• Self-supervised & Representation Learning
• High-Performance Computing for AI
• Distributed & Multi-GPU Training
• Embedded Vision & Edge AI (TinyML)
• End-to-End AI System Design

• Python, C++, Java, MATLAB

• CNN-based image understanding
• Vision Transformers (ViT) and hybrid CNN–Transformer models
• Medical image preprocessing and large-scale patch-based pipelines
• Representation learning with foundation models (e.g. DINOv2)
• Weakly supervised learning (CLAM)
• Model optimization and quantization

• Multi-node, multi-GPU training on HPC clusters
• Distributed data parallel training (e.g. PyTorch DDP)
• Large-scale data loading and preprocessing pipelines
• Slurm-based environments (srun, sbatch)

• ESP32-CAM
• TinyML deployment
• On-device inference and hardware-level control

• PyTorch, TensorFlow
• OpenCV
• Git / GitHub
• Flask (data collection backend)
• Figma (UI/UX design)

• Processed large-scale histopathology datasets using patch-based pipelines
• Built efficient preprocessing workflows for high-resolution whole-slide images
• Trained deep learning models on HPC clusters with multiple nodes and GPUs
• Implemented distributed training to scale medical imaging experiments
• Applied Vision Transformers (ViT) and DINOv2 for representation learning
• Used CLAM for weakly supervised tumor region modeling and analysis

This project reflects my experience in scaling computer vision research using real-world HPC infrastructure.

• Designed an end-to-end embedded vision system on ESP32-CAM
• Captured images via MJPEG streaming from the on-board camera
• Built a Flask-based backend for data collection and dataset management
• Trained CNN models and optimized them for TinyML deployment
• Performed fully on-device inference with gesture-triggered hardware control
• Achieved a standalone Edge AI system without cloud dependency

This project demonstrates my ability to translate vision models into deployable embedded systems.

• Email: zhuyiming040@gmail.com
• GitHub: https://github.com/LouisZhu040
• Linkdin: www.linkedin.com/in/yiming-zhu-574091326