Senior Research Data Engineer

• Join Canva’s multimodal agent research team as a Senior Research Data Engineer in London. • Own the data foundations behind cutting-edge multimodal agentic research. • Build the pipelines, datasets and tooling that turn research ideas into trainable reality. • Manage the full data lifecycle, from collection and curation to delivery into training pipelines. • Enjoy significant autonomy over how data problems are solved. • Work hybrid, with the option to work from home and on the Shoreditch campus.

• Bring strong Python skills and experience building production-grade data pipelines and ML DevOps. • Show practical prompt engineering experience for reliable LLM and VLM outputs. • Bring experience with large-scale ML data processing and loading using Ray or similar. • Understand data versioning, tokenisation, batching and sharding for training. • Show hands-on experience with data pipelines for large-scale distributed training runs. • Know annotation tooling and human-in-the-loop data collection such as Label Studio. • Understand what good data looks like for LLM and VLM fine-tuning. • Load and write large datasets using AWS and distributed storage. • Scope ambiguous problems with researchers and turn needs into clear plans.

• Bring experience with preference data collection for RLHF or reward modelling. • Know multimodal data such as image-text pairs, video and design assets. • Show experience building synthetic data generation pipelines with LLMs. • Bring a background in data quality metrics and monitoring systems. • Show contributions to dataset releases or benchmarks in the ML community.

• Build pipelines for agent training covering collection, filtering, deduplication, formatting and versioning. • Maintain infrastructure for efficient data loading, storage and retrieval at scale. • Turn research requirements into concrete data specifications with research scientists. • Create evaluation datasets and benchmarks that surface real failure modes. • Build tooling for annotation workflows, synthetic data and preference data collection. • Own data quality through validation frameworks and drift and contamination monitoring. • Document datasets thoroughly, including provenance, limitations and versioning. • Build test coverage for data pipelines and ML workflows. • Raise codebase quality through reviews, refactoring and best practices. • Identify data bottlenecks and propose solutions for the team roadmap.

• Work hybrid, with the option to work from home and on the London campus. • Help build the AI that powers Canva from its London team. • Take significant ownership over how data problems are solved. • Work at the cutting edge of multimodal agent research. • Interview entirely virtually.

Canva is on a mission to empower the world to design, building AI that feels magical for millions of people. Its research team explores multimodal agentic architectures, pre- and post-training and design agents, and turns breakthroughs into product features. Canva’s London campus in Hoxton Square is one of the places where its AI gets built.