Problem area
Desktop 3D printing offers a level of customization that consumer goods markets can't match; instead of buying a generic phone case, water bottle holder, or desk stand off Amazon for $10–50, a user could print one made exactly to their needs. But the workflow from idea to printed object requires CAD modelling, slicer configuration, and materials knowledge, which puts it out of reach for most people. The result is that the benefits of personal fabrication are effectively restricted to users with engineering or technical backgrounds, while everyone else defaults to mass-produced purchases.
Main objectives
The project aims to design and build a Conversational 3D Printer: a system that lets a user describe an object in plain language and receive a print-ready file, without touching CAD software.
Specifically, the project will:
- Develop a pipeline that converts natural-language descriptions into parametric 3D geometry and exports validated, printable STL files.
- Use retrieval-augmented generation to ground outputs in print-feasibility constraints such as tolerances, overhangs, and material limits.
- Evaluate whether non-technical users can successfully produce usable objects through the system.
The broader objective is to make personal fabrication a practical alternative to online purchasing for small everyday items.
Scope of work
The project will begin with a literature review covering text-to-3D generation, LLM tool use, and design-for-additive-manufacturing constraints, followed by an environmental scan of existing text-to-CAD and generative design tools to identify gaps. The team will then define a scoped object domain (simple functional household items) and build a prototype pipeline connecting an LLM front end, a RAG knowledge base of printing constraints, a geometry generation or parametric-CAD backend, and automated mesh validation. The prototype will be tested by printing generated models on physical hardware and iterating on failure cases. Finally, a user study with participants of varying technical backgrounds will assess usability and output quality, with results documented in a final report and demonstration.
Deliverables
- Report
- Presentation
- Prototype
Frequency
- TBD
Skills and training required
The project draws on software development skills in Python, including API integration, prompt engineering, and RAG system design, as well as familiarity with computational geometry and parametric CAD scripting. Practical 3D printing experience is needed to diagnose print failures and encode manufacturability constraints. Research and analytical skills are required for the literature review, experimental design, and evaluation of system outputs, and human-computer interaction and user study methods are needed to run and analyze the usability evaluation. Technical writing and presentation skills round out the requirements for communicating results.
Resources required
The project requires access to an FDM 3D printer and filament for iterative physical testing, along with standard slicing software such as PrusaSlicer or Cura. On the computing side, it needs API access to commercial LLMs (or GPU compute if self-hosting open-weight models), a vector database for the RAG component, and CAD/mesh libraries such as CadQuery, OpenSCAD, or Blender's Python API for geometry generation and repair. Additional needs include cloud or lab compute for experiments, and REB approval plus participant compensation if a user study is conducted.