Join the team redefining how the world experiences design.

Hey, g'day, mabuhay, kia ora,你好, hallo, vítejte!

Thanks for stopping by. We know job hunting can be a little time consuming and you're probably keen to find out what's on offer, so we'll get straight to the point. 

Where and how you can work

Our head office is in Sydney, Australia, but San Francisco is now home to our US operations. The role is listed as hybrid, meaning we are flexible and empower you to work where you prefer - whether that's at home or at the office. 

About the Role:

At Canva, we’re building a future powered by AI that’s as magical as it is impactful. As a Senior Research Scientist (Generative Video), you’ll help push the boundaries of video generation and editing—turning cutting-edge research into practical, scalable capabilities that empower millions of creators.

This role blends hands-on applied research with strong technical ownership. You’ll design, train, and evaluate generative video models, collaborate closely with engineering and product partners, and help translate breakthroughs in video diffusion and multimodal learning into real-world experiences in Canva.

At the moment, this role is focused on:

• Own and deliver research projects that advance Canva’s generative video capabilities (text-to-video, image-to-video, video-to-video, video editing).

• Design and run rigorous experiments to validate hypotheses, improve quality, controllability, temporal coherence, and runtime performance.

• Develop and improve model architectures and training pipelines for video generation, including diffusion-based approaches and complementary techniques.

• Translate research into production impact by partnering with ML engineers to scale training/inference and integrate models into Canva’s product ecosystem.

• Advance evaluation and benchmarking for generative video, including perceptual quality, motion fidelity, temporal consistency, identity preservation, prompt adherence, safety, and robustness.

• Explore data strategies for video (curation, filtering, deduplication, captioning/annotation, synthetic data, bootstrapped labeling) that improve model reliability and controllability.

• Contribute to the research roadmap by tracking emerging trends, proposing new directions, and identifying high-leverage problems in generative video.

• Share knowledge through internal write-ups, talks, cross-team reviews, and (where appropriate) external publications or conference engagement.

You’re probably a match if you:

You thrive in ambiguity, love connecting deep research to product outcomes, and can independently drive meaningful research work from idea to deployment. You balance scientific rigor with practical delivery, communicate clearly with cross-functional partners, and have strong instincts for what will make models useful.

We’re looking for someone who brings:

• Deep expertise in generative video modeling, including strong familiarity with modern approaches such as:

• Video diffusion (latent diffusion for video, spatiotemporal U-Nets/DiTs, conditional diffusion, guidance strategies, scheduler choices).

• Temporal modeling techniques (3D/2+1D convs, temporal attention, factorized attention, optical-flow-aware modeling, recurrent/streaming approaches).

• Controllability methods (ControlNet-style conditioning for video, pose/depth/segmentation conditioning, motion control, camera control, keyframes, masks, and edit constraints).

• Consistency and identity preservation (subject-consistent generation, reference-based conditioning, feature/embedding locking, token/adapter strategies, multi-view constraints where relevant).

• Efficient training and adaptation (LoRA/adapters, distillation, latent-space tricks, progressive training, multi-stage pipelines, mixed precision, distributed training).

• Longer-horizon video generation strategies (hierarchical generation, chunked/overlapped sampling, latent caching, frame interpolation, consistency models, or hybrid autoregressive + diffusion pipelines).

In addition, you have:

• Experience developing and deploying generative AI systems (video synthesis/editing strongly preferred; multimodal systems also valuable).

• Strong working knowledge of multimodal representation learning (video-text, video-image, VLM-style conditioning, retrieval-augmented conditioning).

• A solid publication record or demonstrable research impact in industry (shipped systems, patents, open-source contributions, or measurable product gains).

• Experience taking a research problem end-to-end

• Proficiency turning complex papers/ideas into robust implementations and evaluating them with scientific rigor.

• A collaborative mindset: you partner exceptionally well with engineering, product, design, and other researchers.

Bonus points (nice to have)

• Experience with video editing models (inpainting/outpainting, temporal-aware masking, object removal, background replacement, stylization, relighting).

• Familiarity with video safety and responsible gen-AI practices (content filtering, watermarking, provenance, bias, prompt abuse mitigation).

• Experience building human + automated evaluation loops for generative video quality and preference optimization (reward models, RLHF-style tuning, DPO variants, or preference learning).

• Understanding of inference optimization (quantization, compilation, batching, KV cache strategies, streaming generation, GPU memory optimization).

What you’ll learn and how you’ll grow

• Deep involvement in Canva’s long-term AI strategy for generative media and multimodal systems.

• Opportunities to mentor and strengthen our research culture through reviews, best practices, and technical leadership.

• Exposure to product impact at global scale (hundreds of millions of users).

• Dedicated time for self-driven research exploration and experimentation, with strong pathways to ship outcomes.