SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
inDrive is building the data foundation for autonomous-driving work. You will own the design and operation of multi-sensor data pipelines that ingest, synchronize, and geo-reference data from vehicles equipped with cameras, IMU, LiDAR, and GPS sensors. Your work will transform raw sensor streams into training datasets for autonomous-driving models.
Key responsibilities:
- Design and build pipelines to move multi-sensor data (multi-camera video, IMU, LiDAR, GPS) from vehicles into cloud storage, with proper synchronization and geo-referencing
- Own storage architecture and lifecycle management: partitioning strategies, hot/archival tiers, and cost optimization as data volume scales
- Implement validation and monitoring on the ingest path to catch incomplete uploads, missing or unsynchronized streams, and vehicles producing degraded data
- Prepare training datasets for autonomous-driving models: select data by location, route, and environmental conditions; version datasets; deliver in formats ML teams require
- Initially, take over and enhance existing imagery and GPS-track pipelines from vehicle cameras, and contribute to geodata pipelines (ingestion from company data, OpenStreetMap, Overture Maps, and vendors; conflation; validation; publication)
- Optimize spatial queries on large PostGIS and GeoParquet tables in BigQuery
Requirements:
- 4+ years of data engineering experience
- Hands-on experience with multi-sensor data from vehicles or robots (cameras, IMU, LiDAR) and keeping streams synchronized
- Experience preparing training datasets for computer vision or autonomous-driving models
- Proficiency with large volumes of binary/media data in object storage (S3 or GCS) and distributed processing (Spark or equivalent)
- Strong spatial data skills: PostGIS, GeoPandas/Shapely, coordinate reference systems, GPS tracks
- Strong Python and SQL; AWS or GCP; Docker; Kubernetes
Nice to have:
- Sensor data formats (ROS2 bag, MCAP, point clouds)
- Parquet, Delta, or Iceberg; Databricks
- Argo Workflows or Airflow
- OpenStreetMap and Overture Map knowledge
- QGIS
- BigQuery
