PhD Research Scientist Intern

Canva • San Francisco, CA

5d • Hybrid

About The Position

Join the team redefining how the world experiences design. We’re looking for current PhD students ready to bring their research into the real world and help shape the culture of AI at Canva. Our full-time, 16 week AI Research Internship starts in September. During your internship, you’ll work directly with Canva’s AI team on a live, industry-scale project, turning part of your PhD journey into real world impact. You’ll gain hands on experience with real data, production infrastructure and real deadlines, while learning from and working alongside the researchers and engineerings creating Canva’s next generation of AI-powered experiences. At Canva, we're building a future powered by AI that's as magical as it is impactful. As a Generative AI Research Scientist Intern at Canva, you'll advance the frontier of agentic AI—building autonomous systems that can reason, plan, use tools, and act on behalf of hundreds of millions of users. You'll explore the boundaries of what agents can achieve in real-world productivity and creative scenarios, and bring that research into Canva's core product experiences—across design, presentations, documents, spreadsheets, video, and other multimodal creation and collaboration workflows.

Requirements

• You’re currently completing a PhD (ideally third year or later).
• You have experience with human-AI interaction, conversational agents, or user-modeling research.
• You can design, run, and interpret machine-learning experiments with strong scientific rigour.
• You can develop research code for data processing, model training, and evaluation.
• You communicate technical work clearly in writing and presentations.
• You enjoy collaborating with researchers and engineers on technically challenging problems.
• You’re a clear communicator who can collaborate effectively across teams.

Responsibilities

• Developing a synthetic data pipeline that produces conversations with our design agent — user turns, tool calls, and design-state changes — conditioned on personas and statistical criteria on requests derived from real usage.
• Combining persona-conditioned user-simulation techniques (e.g. evolved/adversarial personas, dual-control agent frameworks) with our existing intent-labeling and clustering pipeline, to keep synthetic output in-distribution with real usage.
• Measuring realism gap, distributional fit, and downstream evaluation quality (separability and agreement against our existing ELO rankings) with rigorous, reproducible validation.
• Collaborating with the eval/labelling, research, and product teams to determine whether and how synthetic data can substitute for real user-generated content in evaluation.
• Contributing methodology and findings back to the broader research community through publication where results support it — this is an active, largely unsolved area of research.

What This Job Offers

• Job Type: Full-time
• Career Level: Intern
• Education Level: Ph.D. or professional degree