Important: this is the job description provided by the company — we intentionally left it unchanged. You can read our fun texts on Telegram, both in Russian and in English.
Founded in the US in 2022 and now based in London, UK, Recraft is an AI tool for professional designers, illustrators, and marketers, setting a new standard for excellence in image generation.
We designed a tool that lets creators quickly generate and iterate original images, vector art, illustrations, icons, and 3D graphics with AI. Over 3 million users across 200 countries have produced hundreds of millions of images using Recraft, and we’re just getting started.
Join a universe of professional opportunities, develop and support large-scale projects, and shape the future of creativity. We are committed to making Recraft an essential, daily tool for every designer and setting the industry standard. Our mission is to ensure that creators can fully control their creative process with AI, providing them with innovative tools to turn ideas into reality.
If you’re passionate about pushing the boundaries of AI, we want you on board!
At Recraft, we’re building the next generation of generative models across images and text. We’re looking for an ML Data Engineer to scale our data pipelines for unstructured data (primarily images) and keep our training flows fast, reliable, and repeatable. You’ll design and operate high-throughput ingestion and preprocessing on Kubernetes, evolve our internal data-pipeline framework, and work hand-in-hand with ML engineers to ship datasets that move model quality forward.
Develop and maintain data-ingestion pipelines to source and prepare large-scale image (and occasional text/HTML) datasets from open, publicly accessible, and permitted sources.
Own the end-to-end flow: raw data → quality/beauty/relevance filtering → dedup/validation → ready-to-train artifacts.
Operate and improve our Kubernetes-based data-pipeline framework (distributed jobs, retries, monitoring, automation).
Work with S3-style object storage: efficient layouts, lifecycle, throughput, and cost awareness.
Add tooling around pipelines (progress/health visualization, metrics, alerts) for observability and faster iteration.
Collaborate closely with ML engineers to align datasets with training needs and accelerate experimentation.
Must-have