Research Engineer - Design Generation Modeling

Location: Team Engineering

Team: Country San Francisco / United States

About the Role

At Canva, we're on a mission to empower everyone in the world to design anything and publish anywhere. As a Research Engineer in Design Generation Modeling within our Engineering team, you'll collaborate with brilliant researchers across the globe to push the boundaries of generative design AI. This onsite role in San Francisco focuses on building foundation models for design understanding and generation, visual models for design creation and decomposition, and innovative design representations for discrete modeling with multimodal language models. You'll develop end-to-end experimental training pipelines—from data curation to scalable inference—delivering artifacts that redefine creative possibilities for millions of Canva users. Your primary impact will come from optimizing ML data pipelines, enhancing training workflows, and partnering with research scientists on cutting-edge experiments. Dive deep into complex challenges, leveraging distributed training frameworks like DeepSpeed and FSDP, while staying current with GANs, diffusion models, transformers, and vision-language advancements. With disciplined coding practices and cloud expertise (AWS ideal), you'll improve our internal codebases and prototype compelling ML demos that bridge research and product magic. Our collaborative, innovative culture thrives on passionate builders who love both fundamental research and user-focused outcomes. This is your chance to join a team that's not just building models, but reimagining design at global scale. Enjoy Canva's supportive environment with equity, health benefits, 401(k) matching, inclusive parental leave, flexible options, and that signature Canva vibe of connectivity and fun. Interviews are virtual, and we hire based on skills, experience, and cultural fit—apply even if you don't match every bullet. Let's create the future of design together!

Key Responsibilities

• Develop and optimize ML data pipelines and training workflows for generative design models.
• Implement experimental training pipelines from data preparation to inference.
• Improve internal training codebases and engineering procedures.
• Collaborate with global research scientists to design and execute experiments and ablation studies.
• Contribute to foundation models for design understanding and generation.
• Build visual models for design generation and decomposition.
• Develop innovative design representations for discrete modeling with multimodal language models.
• Dive deep into complex engineering problems to push state-of-the-art boundaries.
• Prototype ML demos with user-friendly interfaces to showcase research impact.
• Conduct code reviews and maintain high coding standards in a collaborative environment.
• Work in cloud environments to scale training infrastructure efficiently.
• Bridge research and product teams to deliver impactful generative design capabilities.

Required Qualifications

• Experience with distributed training at scale using libraries like DeepSpeed, FSDP, Torch, Titan, etc.
• Familiarity with the literature around GANs, diffusion modeling, transformer architectures, and vision language modeling.
• Experience working in cloud environments, ideally AWS.
• Disciplined coding practices with experience in code reviews and pull requests.
• Strong understanding of ML data pipelines, training workflows, and experimental design.
• Passion for both product-focused and basic research in generative AI.
• Ability to collaborate effectively with research scientists and ML engineers.
• Bachelor's or higher degree in Computer Science, Machine Learning, or related field.

Preferred Qualifications

• Specific experience with modeling design data.
• Experience with Graph Neural Networks.
• Ability to quickly prototype ML demos with appealing user interfaces (e.g., Gradio, custom interfaces).
• Prior work on foundation models for design understanding and generation.
• Experience with visual models for design generation and decomposition.
• Background in developing design representations for discrete modeling with MLMs.

Required Skills

• Distributed training (DeepSpeed, FSDP, Torch, Titan)
• GANs and diffusion modeling
• Transformer architectures
• Vision language modeling
• ML data pipelines and training workflows
• Cloud computing (AWS preferred)
• Code reviews and pull requests
• Experimental design and ablation studies
• Python and PyTorch proficiency
• Problem-solving in complex engineering contexts
• Collaboration with research and engineering teams
• Prototyping ML demos
• Design data modeling
• Graph Neural Networks
• Multimodal foundation models
• Inference optimization
• Agile development practices
• Clear technical communication

Benefits

• Equity packages to share in Canva's success
• Comprehensive health benefits plans supporting your wellbeing
• 401(k) retirement plan with generous company contribution
• Inclusive parental leave policy for all parents and carers
• Annual Vibe & Thrive allowance for wellbeing, social connection, and office setup
• Flexible leave options to recharge and support personal needs
• Hybrid work flexibility empowering you and your team
• Regular moments of magic, connectivity, and fun woven into Canva life
• Opportunities to join a team redefining how the world experiences design

Canva is an equal opportunity employer.