Senior Data Platform Architect

• Proven commercial experience in data architecture or data engineering on platforms at scale, with hands-on delivery.
• Experience with images, video or point clouds on object storage, backed by metadata catalogs and registries.
• Practical experience with curation, dataset splits, holdouts and reproducible dataset versions.
• Hands-on lineage and versioning from raw data to training datasets using LakeFS, DVC, Delta or equivalent.
• Strong knowledge of Parquet, Iceberg or Delta, plus batch and streaming ingestion with Kafka or Kinesis.
• Hands-on work with at least one major cloud (AWS S3, Azure or GCP), Spark or Databricks, and Python.
• Comfort using AI coding tools to turn architecture into working software quickly.

Nice to have:

• Experience with autonomous vehicle, robotics or ADAS data (ROS bags or MCAP, sensor logs, drive or scene data).
• Experience building annotation and labeling pipelines, including auto-labeling.
• Work with synthetic or simulated imagery and sensor data for training or testing.
• Hands-on use of MLflow or similar tools for experiment and artifact tracking.

• Curation & Annotation Flow: Design the end-to-end flow from landed sensor data to training-set creation for a data-collection fleet scaling to 500+ machines.
• Metadata Registry & Cataloging: Build the metadata registry for every asset and define the cataloging approach used by ML and robotics engineers.
• Lineage, Versioning & Lifecycle: Set up lineage, versioning, lifecycle and retention from raw sensor logs to each training dataset, so every model can be traced back to its source data.
• Simulation-Ready & Synthetic Data: Define how real data becomes simulation-ready and how synthetic data is stored, versioned and mixed with real data.
• Data Contracts & Implementable Architecture: Agree data contracts with the perception and kit teams, and produce architecture artifacts engineers can start building on right away, working hands-on with AI coding agents.

• Willingness to travel to client sites when needed.