PhD Intern - AI Researcher for CAD (B-Rep) Generation

Job Requisition ID #

25WD92254

Position Overview

The intern will join our team of scientists at the Autodesk AI Lab, working at the intersection of generative AI and computer-aided design (CAD). They will help investigate how large language and vision models (V/LLMs) can be adapted to understand and generate CAD objects and assemblies represented in boundary-representation (B-Rep) formats, with an emphasis on producing parametric, editable, and manufacturable designs. Reporting to a Senior Principal AI Scientist, the intern will work on finetuning LLMs for very long-context sequences and exploring techniques such as external memory, retrieval-augmented generation (RAG), and tool calling. A key focus will be enabling models to produce high-quality outputs that rely on accurate internal "references," and cross-linked identifiers within CAD structures.

Responsibilities

• Finetune LLMs such as Qwen 3, gpt-oss to generate CAD shapes.
• Develop techniques based on RAG and external memory to handle long contexts of over 30,000 tokens that may contain CAD models that have been previously generated or exist in catalogues.
• Evaluate model outputs using standard metrics for CAD shapes including validity of the B-Reps, and prompt following. - Document research findings, present internally to stakeholders and publish if new discoveries are made.

• Currently pursuing a Master's or PhD in Computer Science, Artificial Intelligence, Machine Learning, or a related field
• Strong understanding and hands-on experience with deep learning and LLMs, particularly, experience in finetuning large frontier models, tool calling, and RAG
• Strong Python coding skills and collaborative development using Git
• Familiarity with 3D topological data structures such as the half-edge data structure meshes and B-Reps, geometry processing, and CAD frameworks such as OpenCascade is a strong advantage