Senior Machine Learning Engineer - Agent Tools Interop (AU remote)

About the Group/Team: The Ecosystem Supergroup is focused on creating an abundant app ecosystem that tailors Canva for over a billion delighted users. Our mission is to empower developers and partners to extend Canva in powerful ways, unlocking infinite creative workflows and integrations across the internet. Within Ecosystem, the Apps APIs group builds the developer platform that allows partners to build apps and integrations directly inside Canva. We create the infrastructure, APIs, SDKs, and tooling that enable developers to extend Canva’s capabilities and deliver new creative experiences to millions of users worldwide. About the Role/Specialty: As a Senior Machine Learning Engineer on the Agent Tools Interop team, you'll work at the frontier of agentic AI, specifically on the problem of making tool calling reliable, safe, and scalable across Canva's growing ecosystem of agents. The core challenge: designing systems where agents reliably find and invoke the right capability, at scale. That requires tool definitions models can reason over, frictionless contribution for internal teams, and evaluation infrastructure to keep invocation quality production-grade. Your work accelerates every team at Canva building agentic features. The interoperability layer you help build becomes the shared foundation they depend on, eliminating bespoke wiring and giving every new AI integration a governed, reliable place to plug in. You'll work closely with GenAI teams, product, and platform engineering, bridging ML thinking with systems design to make agentic tool use something Canva can depend on at scale.   What you’ll do (responsibilities): Build and evolve the systems that enable agents to discover, invoke, and safely execute capabilities across Canva at scale, from initial foundations through to long-term platform maturity. Design tool schemas and definition patterns that maximize LLM tool selection accuracy and reliable invocation across diverse agent consumers and AI integrations. Build and operate evaluation pipelines that measure tool calling behavior in production, catch regressions, and drive continuous quality improvement. Collaborate with product, platform, and GenAI teams to integrate agentic capabilities into production systems and understand how tool use behaves at real-world scale. Advise contributing teams on how to define tools agents can reliably call, lowering the bar for onboarding new capabilities into the shared agentic layer. Partner with platform engineers on governance, safety, and execution guarantees, so the platform earns trust as a company-wide dependency for AI integrations. Stay close to developments in agentic AI (MCP, function calling, A2A protocols) and apply new approaches where they meaningfully improve tool interoperability. Mentor engineers on agentic integration patterns and evaluation methodology.   What we're looking for: You have hands-on production experience with LLM tool-use and function calling: designing tool schemas, understanding how model behavior is shaped by tool definitions, and shipping agentic integrations to real users. You've built evaluation frameworks that measure AI feature quality systematically, and you use those signals to drive improvement rather than just report them. Java proficiency is essential given our backend services infrastructure; Python or TypeScript is a strong plus. What sets the strongest candidates apart is experience at the boundary of ML and platform engineering, where you've had to collaborate deeply with backend or infrastructure teams to make AI integrations production-grade, safe, and scalable. Familiarity with MCP, LangChain, LangGraph, or agent frameworks is a real differentiator, as is prompt engineering experience specifically for tool definitions and tool calling schemas. This role is not a fit for research-only backgrounds, traditional ML without LLM/GenAI exposure, or data engineering experience presented as ML engineering ... (truncated, view full listing at source)