Senior Research Engineer - Video Agents

About the role

In your role as Senior Research Engineer, you'll be at the heart of building the next generation of generative video and audio technology. Your work will push the boundaries of what's possible in AI-driven storytelling, helping Canva users go from idea to finished video with tools that feel genuinely magical. By developing and shipping cutting-edge video generation models, you'll help redefine how hundreds of millions of people create and tell stories visually, shaping one of the most exciting product frontiers in AI.

About the team

Our Video Storytelling team builds the agentic systems behind Canva's video product. We sit between our generative research teams and the millions of creators who use Canva to tell stories every day. Our job is to make video creation feel more like directing and less like editing.

We're hiring a Research Engineer to help define how Canva's video agents think, plan, and ship.

You'll work hands-on across the agent stack, from planning and tool orchestration to memory, evals, and shipping. You'll partner closely with product, design, and our generative-models teams to turn research into reliable creator experiences.

What you'll do

• Design the planning layer of our video agent: how it breaks a creative brief into a sequence of generation, editing, and asset decisions.

• Build and evaluate tool-use behaviours: how the agent chooses among video models, editing primitives, retrieval over a creator's assets, and external services, and how we measure when it chose well.

• Own context and memory for multi-turn video sessions, including long-running projects, brand context, and creator preferences.

• Drive the evals flywheel. Build offline and online evaluations that capture narrative quality and intent satisfaction, not just model metrics.

• Ship to production with the product team. Move research prototypes into Canva features at a pace that keeps up with the model frontier.

• Push the multimodal frontier for storytelling across script, image, video, audio, and timeline structure.

What you bring

• 5+ years of ML research or research engineering experience, with a track record of shipping LLM-powered systems to production.

• Hands-on agent-building experience in at least one of: tool-use architectures, planning loops, memory and context management, or agent evals.

• Strong Python and PyTorch. Comfortable in modern agent frameworks.

• Experience with multimodal models, and a clear view of where they break inside larger systems.

• Strong experimental instincts: tight baselines, clean ablations, data-backed conclusions.

Nice to have

• Prior work on creative tools, video editing, or design tooling. Bonus if you're a regular user of creative tools yourself.

• Familiarity with diffusion and video generation models, as inputs to an agent rather than as a trainer.

• Experience designing evals for LLM-powered products: LLM-as-judge, rubric-based eval, online quality metrics.

• Open-source or published work in agent systems, multimodal reasoning, or video understanding.