My Engineering Daybook

A step-by-step walkthrough of building a digit classifier using the MNIST dataset, starting from raw pixel math, building up through PyTorch layers, and finishing with all 10 digits at 99% accuracy using ResNet18 and fastai.

Building a multi-class image classification model using fastai to identify four types of game controllers (PS5, Xbox Series X, Nintendo Switch Pro, and GameCube). Covers data collection, cleaning, fine-tuning ResNet18, and achieving 97.4% validation accuracy with minimal data.

My thoughts, decisions, and the problems I ran into while building a micro-services-based file storage system using Bun, AWS S3, and Qwik.