Staff Machine Learning Engineer - Integrations & Solutions Group (AU remote)

About the team We aim to build the world’s richest Ecosystem of apps for visual design and communications that supercharge Canva’s ability to meet the diverse needs of a billion users and drive Canva’s MAU growth. With the proliferation of innovations in generative AI and as large organizations have increasingly complex workflows to manage their designs, Canva is well poised to become the platform that simplifies access and discovery of services related to visual design. We are already working with hundreds of developers to enable access to these exciting products with our Apps SDK and Canva public REST API! As one of the groups in Ecosystem, the Integrations and Solutions Group builds Canva's AI integration layer — the infrastructure that connects Canva's design capabilities to the world's leading AI assistants. Our teams own the MCP platform, AI-optimised tools and APIs, evaluation frameworks, and the agent architecture that powers millions of AI-driven design interactions. We work directly with partners like OpenAI, Anthropic, and Google, and we're in the middle of a strategic shift from protocol-first integrations toward intelligent, Canva-hosted design agents. As a Staff Machine Learning Engineer, you'll play a critical role in defining how AI platforms experience Canva — driving the technical standards, evaluation frameworks, and architectural decisions that ensure our integrations are reliable, measurable, and evolving with a rapidly changing ecosystem. What you’d be doing in this role As Canva scales change continues to be part of our DNA. But we like to think that's all part of the fun. So this will give you the flavour of the type of things you'll be working on when you start, but this will likely evolve. At the moment, this role is focused on: Drive the design and evolution of AI-ready tools and APIs that enable LLM platforms (ChatGPT, Claude, Gemini and others) to reliably interact with Canva's design capabilities — defining the patterns and standards that other teams adopt for tool descriptions, payload structures, and intent-based interfaces. Pioneer agent-to-agent communication approaches. Own and evolve evaluation frameworks that systematically measure tool-use accuracy across platforms — defining what "good" looks like for proxy-based fast evals and real-client production evals, and ensuring these frameworks scale as we add platforms and capabilities. Shape Canva's agent architecture — making strategic technical decisions about where intelligence should live (in external LLMs vs Canva-hosted agents), building the orchestration layers that allow third-party providers to invoke Canva's design tools at scale, and driving automation of complex workflows like marketing campaigns. Define and build observability systems that give multiple teams visibility into how AI assistants consume Canva's tools in production — identifying failure patterns, setting quality benchmarks, and closing the loop between production data and continuous improvement. Work across team and platform boundaries — proactively identifying problems not yet defined, understanding behavioural quirks across LLM platforms, and driving solutions that span the AI Integrations, API Capabilities, and Workflow Integrations teams. Influence integration strategy with AI partners and internal teams — contributing to technical direction with OpenAI, Anthropic, and Google, and shaping how Canva's broader engineering organisation builds for AI consumption. You're probably a match if You've shipped LLM-powered systems into production and can quantify your impact — things like improving tool-call accuracy from X% to Y%, reducing agent error rates, cutting latency by Z%, or measurably improving user outcomes through better AI integration. We care about results, not just launches. You proactively find and solve problems others haven't defined yet — you've identified gaps in how LLMs consume tools or APIs, scoped the problem across teams, and driven t ... (truncated, view full listing at source)