Description
Luma is hiring a research scientist to build and train large-scale multimodal agentic models that reason, plan, code, and call tools. The role spans modeling, data pipelines, GPU-scale training, and evaluation, requiring strong foundation-model and agentic-systems knowledge, PyTorch, and large-scale training experience. The position is described as a multi-stack research role with no existing playbook and is intended for someone who can work across several layers of the stack.
