AI-Powered Steel Structure Automation

I want to build an automation-building tool that has artificial intelligence at its core and is tailored specifically to steel structures. The goal is to move repetitive, rules-based steps—everything from preliminary sizing through connection detailing—into a single application that learns from past projects and keeps improving its own recommendations.

Here’s the big picture of what I need:

• The AI layer must analyse inputs such as span, load combinations, regional code requirements and preferred member libraries, then generate an initial steel frame layout automatically.
• Users should be able to override choices, and the system has to retrain itself on those corrections so the next project starts closer to the mark.
• Code-checking and simple finite-element verification should happen in the background, flagging any members that do not meet strength or serviceability criteria.
• A clean, modern interface (web or desktop) is essential for quick data entry and for exporting results to industry-standard formats like IFC, DWG or even a straight Bill of Materials CSV.
• I prefer Python for the back end because of its strong AI libraries (TensorFlow / PyTorch), but I’m open to alternatives if you make a convincing case.

Acceptance criteria
1. A working prototype that can create a basic portal frame in structural steel, complete with sizing and code check, from only span, bay spacing, and load inputs.
2. Evidence of learning: after manually adjusting at least five member sizes, the next generated frame should reproduce 60 % or more of those changes without intervention.
3. Export files open correctly in Tekla Structures or equivalent BIM package without data loss.

If you have experience blending structural engineering logic with machine-learning workflows, let’s talk about how quickly we can move from concept to a usable MVP.