Senior Machine Learning Engineer - Multimodal Data

Canva · Enterprise · London, United Kingdom +1 · Information Technology

Canva is seeking a Senior Machine Learning Engineer to own the data lifecycle for their multimodal agent research. This role involves designing and building data pipelines, infrastructure for data processing and retrieval, and tooling for dataset construction, including human annotation and synthetic data generation. The engineer will collaborate with researchers to define data needs, ensure data quality, and contribute to the development of scalable training and evaluation loops for multimodal agentic systems.

What you'd actually do

Design and build data pipelines for agent training: collection, filtering, deduplication, formatting, and versioning across text, image, and multimodal sources.
Build and maintain infrastructure for efficient data loading, storage, and retrieval at scale (S3, distributed systems, streaming pipelines).
Collaborate with research scientists to translate research requirements into concrete data specifications, and iterate as experiments reveal new needs.
Create evaluation datasets and benchmarks in collaboration with researchers—curating task distributions that surface real failure modes.
Develop tooling for dataset construction—including human annotation workflows, synthetic data generation, and preference data collection for RLHF/DPO-style training.

Skills

Required

Python
production-grade data pipelines
ML DevOps
prompt engineering
LLM/VLM outputs
ML data workflows
large-scale data processing and loading
data versioning
format considerations for training
large-scale distributed ML training runs
annotation tooling
human-in-the-loop data collection
ML training requirements
loading and writing large datasets to/from cloud infrastructure (AWS)
distributed storage systems
communication skills
collaborative approach
ownership
iterating quickly

Nice to have

Ray
Label Studio
preference data collection for RLHF or reward modelling
multimodal data (image-text pairs, video, design assets)
synthetic data generation pipelines using LLMs
data quality metrics and monitoring systems
Contributions to dataset releases or benchmarks in the ML community

What the JD emphasized

multimodal agent research
data foundations
training pipelines
datasets
tooling
scalable training and evaluation loops
multimodal agentic systems
data lifecycle
collection and curation
preprocessing
quality assurance
delivery into training pipelines
design and build the systems
reliably and at scale
significant autonomy
data problems
aligning on what problems matter most
agent training
text, image, and multimodal sources
efficient data loading, storage, and retrieval at scale
distributed systems
streaming pipelines
translate research requirements into concrete data specifications
iterate as experiments reveal new needs
evaluation datasets and benchmarks
task distributions
real failure modes
dataset construction
human annotation workflows
synthetic data generation
preference data collection
RLHF/DPO-style training
data quality
validation frameworks
monitor for drift and contamination
establish standards
datasets trustworthy and reproducible
Document datasets thoroughly
provenance
known limitations
intended use cases
versioning history
comprehensive test coverage
data pipelines and ML workflows
reliability
catching regressions early
Elevate codebase quality
code reviews
refactoring
establishing engineering best practices
research velocity scale sustainably
team roadmaps
identifying data bottlenecks
proposing solutions
unblock research velocity
Strong software engineering skills in Python
production-grade data pipelines
ML DevOps
prompt engineering
designing, testing, and refining prompts
reliable LLM/VLM outputs
ML data workflows
large-scale data processing and loading
Ray, or similar
data versioning
format considerations for training
tokenization, batching, sharding
data pipelines for large-scale distributed ML training runs
annotation tooling
human-in-the-loop data collection
Label Studio or internal systems
ML training requirements
good data
LLM/VLM fine-tuning
anticipate downstream issues
loading and writing large datasets
cloud infrastructure (AWS)
distributed storage systems
Strong communication skills
work with researchers
scope ambiguous problems
translate needs into actionable plans
collaborative approach
comfortable taking ownership
iterating quickly
preference data collection for RLHF or reward modelling
multimodal data
image-text pairs
design assets
synthetic data generation pipelines using LLMs
data quality metrics and monitoring systems
Contributions to dataset releases or benchmarks in the ML community

Other signals

multimodal agent research
data foundations
training pipelines
datasets
tooling

Read full job description

Company Description

Join the team redefining how the world experiences design

Hiya, 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

The buzzing Canva London campus features several buildings around beautiful leafy Hoxton Square in Shoreditch. While our global headquarters is in Sydney, Australia, London is our HQ for Europe, with all kinds of teams based here, plus event spaces to gather our team and communities.

You'll experience a warm welcome from our Vibe team at front of house, amazing home cooked food from our Head Chef and a variety of workspaces to hang out with your team mates or get solo work done. That said, we trust our Canvanauts to choose the balance that empowers them and their team to achieve their goals and so you have choice in where and how you work.

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 Machine Learning Engineer to own the data foundations that power our multimodal agent research—building the pipelines, datasets, and tooling that turn ambitious research ideas into trainable reality.

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.

About the role

You'll be responsible for the data lifecycle that fuels our agent research: from collection and curation through to preprocessing, quality assurance, and delivery into training pipelines. You'll work closely with research scientists to understand what data is needed, then design and build the systems to make it happen—reliably and at scale. You'll have significant autonomy over how data problems get solved, while aligning on what problems matter most with the broader team.

What you'll do

Design and build data pipelines for agent training: collection, filtering, deduplication, formatting, and versioning across text, image, and multimodal sources.
Build and maintain infrastructure for efficient data loading, storage, and retrieval at scale (S3, distributed systems, streaming pipelines).
Collaborate with research scientists to translate research requirements into concrete data specifications, and iterate as experiments reveal new needs.
Create evaluation datasets and benchmarks in collaboration with researchers—curating task distributions that surface real failure modes.
Develop tooling for dataset construction—including human annotation workflows, synthetic data generation, and preference data collection for RLHF/DPO-style training.
Own data quality: build validation frameworks, monitor for drift and contamination, and establish standards that make datasets trustworthy and reproducible.
Document datasets thoroughly: provenance, known limitations, intended use cases, and versioning history.
Implement comprehensive test coverage for data pipelines and ML workflows, ensuring reliability and catching regressions early.
Elevate codebase quality through code reviews, refactoring, and establishing engineering best practices that help research velocity scale sustainably.
Contribute to team roadmaps by identifying data bottlenecks and proposing solutions that unblock research velocity.

You're likely a match if you have

Strong software engineering skills in Python, with experience building production-grade data pipelines and ML DevOps.
Practical experience with prompt engineering—designing, testing, and refining prompts for reliable LLM/VLM outputs.
Experience with ML data workflows: large-scale data processing and loading (Ray, or similar), data versioning, and format considerations for training (tokenization, batching, sharding).
Hands-on experience working with data pipelines for large-scale distributed ML training runs.
Familiarity with annotation tooling and human-in-the-loop data collection (Label Studio or internal systems).
Understanding of ML training requirements—you know what "good data" looks like for LLM/VLM fine-tuning and can anticipate downstream issues.
Experience loading and writing large datasets to/from cloud infrastructure (AWS) and distributed storage systems.
Strong communication skills: you can work with researchers to scope ambiguous problems and translate needs into actionable plans.
A collaborative approach, comfortable taking ownership and iterating quickly.

Nice to have

Experience with preference data collection for RLHF or reward modelling.
Familiarity with multimodal data (image-text pairs, video, design assets).
Experience building synthetic data generation pipelines using LLMs.
Background in data quality metrics and monitoring systems.
Contributions to dataset releases or benchmarks in the ML community.

Qualifications

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 range 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, 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.

We celebrate all types of skills and backgrounds at Canva so even if you don’t feel like your skills quite match what’s listed above - we still want to hear from you!

Please note that interviews are conducted virtually.