AI Research Scientist (Multimodal)

TL;DR

AI Research Scientist (Multimodal): Advancing frontier generative AI models for text-to-image, video, and 3D generation with an accent on large-scale pre-training and model architecture innovation. Focus on designing experiments, optimizing training methodologies on high-end GPU clusters, and translating research into production-ready features for millions of users.

Location: Must be based in Beijing, China

Company

Canva is a global visual communication platform empowering everyone to design anything and publish anywhere.

What you will do

• Determine and lead research initiatives aligned with frontier multimodal AI trends and product goals.
• Design and execute large-scale experiments to validate hypotheses on high-end GPU clusters.
• Create novel generative model architectures, training methodologies, and RL algorithms.
• Develop and maintain synthetic data generation pipelines to improve model performance.
• Interpret experimental results and translate them into actionable insights for product development.
• Collaborate with cross-functional teams and share knowledge through internal documentation and external research publications.

Requirements

• Must be based in Beijing.
• Proven experience in developing and training foundational large-scale generative models (diffusion or auto-regressive).
• Strong academic track record with publications in top-tier conferences (CVPR, ICCV, NeurIPS, etc.).
• Extensive experience in large-scale parallel model pre-training across hundreds of GPUs.
• Proficiency in Python, PyTorch, and deep learning frameworks like Transformers, Megatron, and DeepSpeed.
• Ability to communicate complex research findings effectively to both technical and non-technical stakeholders.

Nice to have

• Open-source contributions in the AI/ML domain.
• Experience with cloud computing platforms for model training and deployment.
• Passion for tracking industry trends and emerging AI technologies.

Culture & Benefits

• Opportunity to work on cutting-edge AI research with global impact.
• Access to high-end GPU infrastructure for large-scale experimentation.
• Collaborative environment with experts in the field.
• Open to candidates at all levels, including graduates and interns.