Senior Research Scientist - Reinforcement Learning, MoEs

Company Description

Join the team redefining how the world experiences design.

Servus, 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 flagship campus is in Sydney, Australia but Austria is home to part of our European operations. And you have choice in where and how you work, we trust our Canvanauts to choose the balance that empowers them and their team to achieve their goals.

Fun fact, a big part of our Austrian operations is developing the AI product within Canva to help reimagine how artificial intelligence can be used in design. Pretty cool ha!

Job Description

At Canva, our mission is to empower the world to design. We’re building AI that feels magical and lands real impact for millions of people - helping anyone create with confidence. We’re looking for a senior research scientist who lives and breathes reinforcement learning, agentic systems and mixture of expert models to push the frontier of reasoning, tool use, latency and reliability - and ship it to users.

About the team

We explore multimodal agentic architectures, build scalable training and evaluation loops, and partner closely with product and platform teams to turn breakthroughs into delightful product features. We are a cutting-edge post-training team, developing new multimodal agentic systems. We work on all topics of multimodal modelling, post-training and design agents, we build scalable training and evaluation loops, and partner closely with product and platform teams to turn breakthroughs into delightful product features. We are looking for a person with experience in post-training, reinforcement learning (RL) and mixture of expert models to join our team.

About the role

You’ll drive research directions and play a leading role in hands‑on work across the agent stack—from reward design and policy optimization to planning, memory, and tool orchestration, dataset construction, to post-training, and the development of novel post-training approaches. You’ll design tight experiments, iterate quickly, and land trustworthy conclusions. Most importantly, you’ll help convert research into reliable, safe, and high‑quality product experiences.

What you’ll do

• Develop agent systems (planning, multimodal tool use, retrieval, novel training approaches, modeling ablations) for real tasks in design, vision, and language.

• Scale post-training and RL across distributed systems (PyTorch) with efficient data loaders, tracing/telemetry, stable training of mixture-of-experts (MoE) architectures, and reproducible pipelines; profile, debug, and optimize.

• Contribute to the research agenda for RL/agentic systems aligned with Canva’s product goals; identify high‑leverage bets and retire dead ends quickly.

• Build reward models and learning loops: RLHF/RLAIF, preference modeling, DPO/IPO‑style objectives, offline/online RL, curriculum learning, and credit assignment for multi‑step reasoning.

• Develop simulation and sandbox tasks that surface failure modes (planning errors, tool‑use brittleness, hallucination, unsafe actions) and turn them into measurable targets.

• Help align on rigorous evaluation for agents (task success, reliability, latency, safety, regressions). Stand up offline suites and online A/B tests; favor simple, controlled experiments that generalize.

• Collaborate and ship: work shoulder‑to‑shoulder with product, design, safety, and platform to land research as reliable features—then iterate.

• Share and elevate: mentor teammates, present findings internally, and contribute back to the community when it helps the field and our users.

You’re likely a match if you have

• Depth in implementing and post-training MoEs/LLMs/VLMs/Diffusion models, with a track record of shipped research or publications in MoEs, RL or agents.

• Experience modifying, and adapting open-source models.

• Strong experience with experimental design: tight baselines, clean ablations, reproducibility, and clear, data‑backed conclusions.

• Fluency in Python and PyTorch; you’re comfortable in large ML codebases and can profile, debug, and optimize training and inference.

• Practical experience building agent loops (planning, tool invocation, retrieval, memory) and evaluating multi‑step reasoning quality.

• Hands‑on experience with policy optimization, reward modeling, and preference learning (e.g., RLHF/RLAIF, DPO/IPO, actor‑critic/PPO, offline RL).

• Experience with large‑scale training (distributed training, experiment tracking, evaluation harnesses) and cloud multimodal tooling.

• Experience with RL for MoE architectures.

Nice to have

• Experience with video and audio modelling.

• Experience with multi‑agent settings.

• Strength in alignment and safety evaluations, including red‑teaming and risk mitigation for tool‑using agents.

• Contributions to open‑source, benchmarks, or shared evaluation suites for agents.

Additional Information

What's in it for you?

Achieving our crazy big goals motivates us to work hard - and we do - but you'll experience lots of moments of magic, connectivity and fun woven throughout life at Canva, too. We also offer a stack of benefits to set you up for every success in and outside of work.

Here's a taste of what's on offer:

• Equity packages - we want our success to be yours too
• Inclusive parental leave policy that supports all parents & carers
• An annual Vibe & Thrive allowance to support your wellbeing, social connection, home office setup & more
• Flexible leave options that empower you to be a force for good, take time to recharge and supports you personally

Check out lifeatcanva.com for more info.

Other stuff to know

We make hiring decisions based on your experience, skills and passion, as well as how you can enhance Canva and our culture. When you apply, please tell us the pronouns you use and any reasonable adjustments you may need during the interview process.

Please note that interviews are predominantly conducted virtually.