Research Engineer - Generative Video

Company Description

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

In your role as Staff Research Engineer (Generative Video), you’ll help bring Canva’s next wave of AI-powered video creation to life — turning cutting-edge generative video research into reliable, scalable, production-ready systems that delight hundreds of millions of users.

You’ll sit at the intersection of applied research and engineering, partnering closely with Research Scientists and product engineering teams to shape the end-to-end generative video stack — from data and training, to evaluation, to inference and product integration. This is a hands‑on, Staff‑level role where you’ll set technical direction, make high‑impact trade‑offs, and raise the bar on engineering excellence and operational maturity for generative video at Canva.

At the moment, this role is focused on:

• Working closely with Research Scientists to translate new generative video ideas into practical, scalable implementations (e.g. diffusion‑based video generation, multimodal conditioning, temporal consistency techniques)

• Setting technical direction for generative video projects (text‑to‑video, image‑to‑video, video‑to‑video, and video editing), aligning research bets with product needs, safety expectations, and platform constraints

• Designing and building end‑to‑end training and inference pipelines, evolving prototypes into robust systems with benchmarking, monitoring, regression testing, and production guardrails

• Driving quality and cont rollability improvements through rigorous experimentation — including temporal coherence, identity preservation, prompt adherence, and runtime performance

• Engineering core model + systems components for modern generative video approaches

• Optimizing for scale and efficiency, including distributed training performance, mixed precision, memory/throughput improvements, batching, and system‑level latency/cost trade‑offs in serving

• Advancing evaluation, benchmarking, and data strategy, improving reliability via dataset curation, filtering, deduplication, captioning/annotation, synthetic data, and bootstrapped labeling

• Strengthening operational excellence for production models: observability, incident response, root‑cause analysis, rollbacks, prevention via automated checks and guardrails

• Mentoring and uplifting others through design reviews, code reviews, experiment reviews, and knowledge‑sharing across engineering and research

You’re probably a match if you:

• Thrive in ambiguity and enjoy owning complex, end‑to‑end systems that bridge research and product engineering

• Can make pragmatic trade‑offs between quality, cont rollability, latency, cost, and safety — and bring others along through clear technical communication

• Care deeply about building systems that are not just impressive in demos, but shippable, scalable, and dependable

• Collaborate generously, mentor others, and raise engineering standards wherever you go

We’re looking for someone who brings:

• Strong experience building generative AI systems, ideally in generative video or video editing (multimodal experience is a big plus)

• Solid understanding of modern generative approaches (diffusion models, Transformers/DiTs, GANs) and how they behave in real‑world pipelines

• Strong working knowledge of multimodal learning, including video‑text/video‑image conditioning, VLM‑style conditioning, and/or retrieval‑augmented conditioning

• Staff‑level engineering impact, with a track record of leading technical initiatives across stakeholders — driving alignment, making trade‑offs, and delivering durable outcomes

• Experience scaling training and inference, including distributed training across large GPU fleets…