Cedric Hugo Kuswanto

Handwritten Digit Recognition using C++ and OpenCV

• Utilized OpenCV and classical algorithms (KNN, SVM, Decision Trees, K-means, DNN) for training on MNIST.
• Developed efficient data loading and preprocessing pipeline for image classification.
• Achieved high accuracy on the MNIST test set, demonstrating strong algorithm implementation skills.
• Delivered a comprehensive report with code, results, and performance analysis.

Personal Website Development

• Developed a responsive and interactive personal website using JavaScript, CSS, and HTML.
• Designed and implemented a fully responsive website with a focus on user experience and aesthetic appeal.
• Developed an interactive scrapbook leveraging JavaScript to create a dynamic and engaging user experience.
• Applied CSS for custom styling and responsive layouts to ensure compatibility across different devices.

Pneumonia Detection with PyTorch (In Progress)

• Developing a high-accuracy deep learning model to classify X-ray images into Normal, Bacterial Pneumonia, and Viral Pneumonia using PyTorch.
• Fine-tuning a pre-trained ResNet50 model to leverage transfer learning for improved performance.
• Implementing advanced data augmentation techniques to enhance model robustness.
• Focusing on optimizing the model to achieve high classification accuracy on a medical imaging dataset.

xv6 File System: Memory Mapping and Management

• Enhanced memory management by implementing mmap() and munmap() system calls, significantly improving the efficiency of file access in user space by directly mapping files into process memory.
• Revamped Virtual Memory Area (VMA) structures to support better tracking of memory mappings, improving the system’s handling of memory protection and access rights for mapped regions.
• Optimized page fault handling, dynamically allocating physical memory and seamlessly loading file data into mapped regions, leading to better performance and resource utilization.
• Refined process isolation and memory safety by improving memory protection mechanisms, preventing unauthorized access and ensuring that mappings are handled correctly across different processes.

QTRVSim with RISC-V Vector Extension (RVV) Support

• Extended QTRVSim to support key RVV instructions, enabling efficient vector processing.
• Implemented vector operations including vsetvl, vadd.vv, vadd.vx, vmul.vv, and memory load/store instructions for accelerated computation.
• Achieved significant performance improvements through cycle optimization, reducing execution time for vector operations.
• Developed matrix multiplication using vectors to demonstrate enhanced computational speed and applied advanced reduction techniques for further optimization.
• Maximized simulation efficiency by incorporating techniques like chaining, reducing cycle counts for faster execution.
• Authored extensive test cases to validate functionality and verify performance gains.