Data platform engineer for frontier models

We are seeking a visionary and highly technical Senior Data Platform Engineer to architect, implement, scale, and maintain the data engine powering our next-generation frontier models. In this high-impact role, you will bridge the gap between cutting-edge AI research and high-performance engineering, treating the data platform as an internal product with our researchers as your primary customers.

• Scale and automate the data processing stack to handle petabytes of data and ensure its smooth operation.
• Design the execution layer for pipeline stages of differing computational shape, matching each stage to an appropriate engine and keeping the pipeline saturated end to end.
• Ensure efficient use of compute resources, including GPU access for compute-intensive data processing tasks.
• Make pipelines reliable at scale, with graceful failure recovery, restartability, and observability that attributes bottlenecks and cost to the responsible stage.
• Ensure all datasets, including the intermediate outputs of each transformation stage, are versioned, reproducible, and fully traceable to meet specific and dynamic experiment needs, and are accompanied by datasheets, in accordance with internal Data Governance policies.
• Partner with the Research team to ensure the datasets you produce integrate seamlessly with the training pipelines.

• A bachelor's degree in a relevant field (e.g., computer science, computer engineering, software engineering) is required.
• 5+ years of experience designing, implementing, and managing large-scale distributed data processing systems, or working within large-scale distributed ML data frameworks, with recent experience using e.g. Ray, Apache Spark, workflow orchestrators, Apache Arrow, and/or Parquet.
• Demonstrated ownership of a data processing system under real throughput, reliability, and cost pressure. The specific frameworks matter less to us than evidence that you have had to reason about where a large pipeline breaks and why.
• Experience profiling and optimizing throughput and cost across heterogeneous workloads, including GPU-accelerated stages.
• Experience with dataset versioning, lineage, and reproducibility tooling.
• Ability to collaborate effectively with cross-functional teams, document best practices, and stay updated with the latest advancements in large-scale data processing and software development.
• Experience with workload managers (e.g., Ray, Kubernetes, Slurm).
• Familiarity with containerization tools (e.g., Docker, Enroot).
• Familiarity with data infrastructures and platforms (e.g., vector databases).

• Условий в вакансии нет