LLM Agent Research Scientist 大模型智能体研究科学家
Core
Define and lead research on LLM-powered agents, designing architectures, reasoning methods, tool-use frameworks, and post-training recipes to enable capable, dependable agents for Canva's creative workflows.
Role type
Senior IC LLM Agent Research Scientist
Builds
Autonomous agent systems for design, presentations, documents, video, and multimodal creation workflows
Domain
Generative AI / Agentic AI / Large Language Models
Deliverable
production ML models
Required skills
LLM agent development, post-training, reinforcement learning (RL), reward modeling, synthetic data generation, agent environment design, Python, PyTorch, Transformers, DeepSpeed, Megatron, cloud computing platforms
Preferred skills
Peer-reviewed publications at top-tier AI venues (NeurIPS/ICML/ICLR/ACL/EMNLP/CVPR), impactful open-source contributions, curiosity about emerging AI trends
Technologies
PyTorch, Transformers, DeepSpeed, Megatron, cloud computing platforms
Responsibilities
Design agent architectures and multi-agent frameworks; develop post-training recipes including RL; train models on large-scale datasets using high-end GPUs; create rigorous benchmarks for agent correctness and safety; partner with engineering and product teams to land research capabilities in Canva products
Seniority
Senior, hands-on IC
