This document discusses using machine learning and optimization techniques to design steel connections. It generated a database of over 1,800 steel connection designs varying parameters like plate thickness, bolt size, and loading conditions. It then analyzed the data using regression and neural network models to predict connection performance. It validated the models on a real-world building project, showing the optimized connections improved utilization compared to the original design. Going forward, it aims to further automate steel connection design and incorporate insights from real projects into guidelines and codes.