AI-Driven Rough Diamond Planner

I manufacture natural diamonds and need a purpose-built application that lets my team load a 3D scan of a rough stone and instantly receive the smartest cutting plan. The core of the system should run on Anthropic’s Claude code model, using its reasoning power to compare thousands of orientation possibilities and predict which combination of cuts will deliver the highest total value from each stone.

What the software must do
• Import common 3D scan formats and reconstruct an accurate model of each rough diamond.
• Analyse the model with Claude-powered optimisation logic to suggest and rank multiple cutting schemes aimed at maximising value recovery.
• Let users rotate, zoom, and edit the proposed plans in an intuitive 3D viewer, then lock in the final choice.
• Produce automated reports that show expected yields, carat loss, projected market value, and a step-by-step cutting guide that my workshop can follow.

Please use tools and libraries you already trust for 3D modelling, GPU acceleration, or mesh processing—as long as the AI reasoning layer remains driven by Claude and the end result is a single, streamlined desktop or web application we can run in-house.

Acceptance criteria
1. Upload a sample scan, receive at least three ranked cut plans within five minutes on standard workstation hardware.
2. Visual accuracy of the 3D model within ±0.02 mm of the source scan.
3. Reports exportable in PDF or CSV with clearly labelled metrics.

I’m open to architectural suggestions—cloud versus on-premise, Rust versus Python engines, etc.—so long as the final system is maintainable and the optimisation logic stays transparent (well-documented code, comments, and explanation of Claude prompts). If you have prior experience with diamond planning, gems, or similar 3D optimisation problems, that will weigh heavily in your favour. Let’s create a tool that turns every rough stone into its most valuable finished form.