• Work on image-generation AI models that will be used by more than 135 million users every month and make a global impact.
• Stay on top of the rapidly evolving state-of-the-art within the image generation space and, more generally, in visual AI. Reading papers and evaluating them for our team goals is a key part of your job.
• Study best practices, develop untried approaches, work on prototypes and evaluate their performance.
• Customize and fine-tune open image generation models like Stable Diffusion with technologies such as LoRA, ControlNet, Hypernetworks, and your own ideas.
• Experiment on large-scale cloud infrastructure (GCP and AWS) with high-end hardware, being mostly NVIDIA A10 and A100 Tensor Core GPUs, together with our ML Ops specialists.
• Analyze internal and external data sources, and generate, collect, and prepare data with commercial licensing in mind. Collaborate with our internal data generation team as well as with external partners.
• Set up data augmentation pipelines and perform hyper-parameter searches to build production-ready AI solutions that meet the quality requirements of photo editing effects in Canva.
• Be part of our On-call rosters. At Canva, we believe a healthy on-call culture enables and empowers our engineers to give our users the best possible experience.

You're probably a match if

• You have a Master's (ideally Ph.D.) degree in Computer Science (or similar), focusing on machine learning. Published papers in peer-reviewed journals and conferences would be a valuable addition.
• You have a solid understanding of computer vision, especially the concepts of convolutional neural networks (CNNs), transformers, and diffusion models.
• Proficiency with the following technologies is required to be set up for success in this position: Deep learning frameworks such as PyTorch and Tensorflow (we use PyTorch), Python 3, Numpy, OpenCV, Docker & Kubernetes.
• You have a solid understanding of modern software engineering principles, Python 3 classes and inheritance, multithreading, testing, and QA to write clean code.
• Nice to have: Experience with javascript, node.js, react, typescript, and (WebGL) shaders to build prototype GUIs.
• Our projects require picking up knowledge of the state-of-the-art of a specific ML topic quickly since most projects start with a phase of literature research and communicating the context to the team (across timezones).
• You have an excellent sense of ownership and like to find and develop novel solutions to challenging problems proactively.
• You are an exceptional communicator and team player, eager to work with your colleagues to brainstorm ideas.
• To be employed by Canva for this role, you need to live and work from a country where Canva has an entity set up - you can find all our office locations here.