PhD Intern - Efficient 3D Generation Frameworks

Job Requisition ID #

25WD92246

Position Overview
Are you excited about advancing the future of intelligent 3D content creation? Do you want to explore how learning from feedback can make generative models more efficient, adaptable, and aligned with human intentions? If so, this role is for you.

You will join Autodesk Research to develop efficient 3D generative modeling frameworks that enable rapid sampling and iterative improvement. You will prototype new training approaches that integrate various forms of feedback or guidance, helping models refine and improve their generated output. Working closely with senior research scientists, you'll develop solutions that can influence real-world products such as Fusion and AutoCAD, while contributing publishable insights to the scientific community.
This is an opportunity to help shape the next generation of AI-driven design tools that truly enhance human creativity.

Responsibilities

• Conduct research on accelerated generative modeling for 3D shape synthesis
• Develop feedback-driven learning and refinement algorithms
• Build efficient training pipelines using Python, PyTorch, and distributed frameworks such as Ray or RayLightning
• Prototype and evaluate techniques that improve sampling speed, controllability, and diversity in 3D generation
• Collaborate with Autodesk Research scientists and engineers to integrate methods into real-world design workflows
• Document findings and contribute work suitable for submission to top ML venues (e.g., NeurIPS, ICLR, ICML)

• Currently pursuing a PhD in Computer Science, Machine Learning, Artificial Intelligence, or a related discipline
• Strong foundation in generative modeling, deep learning, and (optionally) reinforcement learning or feedback-driven optimization
• Proficiency in Python and PyTorch
• Understanding of distributed training or experience with Ray/RayLightning
• Strong analytical and communication skills, with the ability to iterate quickly based on experimental results

• Research experience in efficient generative modeling, 3D representation learning, or feedback-based training
• Familiarity with geometric ML, 3D shape representations, or CAD data
• Evidence of research impact (publications or preprints in venues such as NeurIPS, ICLR, ICML, CVPR, etc.)
• Interest in bridging foundational ML research with real-world generative design applications
• Ability to work independently while collaborating across multidisciplinary teams