INNOVATION IS IMAGINING WHAT NO ONE ELSE CAN.
At the BMW Group, everything begins with passion. It transforms a profession into a vocation. It drives us to continually reinvent mobility and bring innovative ideas to the roads. Enthusiasm for collaborative projects turns a team into a strong unit where every opinion is valued. It is only when expertise, highly professional processes, and enjoyment of work come together that we can shape the future collectively.
Responsibilities
- Build and maintain end-to-end ML pipelines using workflow orchestration tools, from data ingestion to distributed training, evaluation, model compilation, and deployment-ready artifacts
- Engineer petabyte-scale data pipelines that consume domain datasets, transforming raw MDF4 and MCAP log files into training-ready formats
- Build tooling for efficient parallel readers, signal extraction, synchronisation of multi-sensor streams, and integration with dataset management platforms for visual QA and curation
- Manage experiment tracking, hyperparameter tuning and model registry, enforcing reproducibility, lineage, and approval gates from experiment to production
- Develop and maintain model compilation and optimisation pipelines targeting in-vehicle Qualcomm Snapdragon Ride chips and/or NVIDIA automotive SoCs
- Operate observability stacks, providing dashboards, data-drift alerts, pipeline SLOs, and log aggregation
Requirements
- University degree in Computer Science, Engineering, or a related field
- 3-5 years of hands-on ML infrastructure or MLOps experience
- Strong Python skills; experience with hermetic build systems (e.g., Bazel) is a plus
- Production Kubernetes experience, including deploying and debugging workloads, writing Helm charts, and managing accelerator node pools
- Working knowledge of ML pipeline orchestration, experiment tracking, and hyperparameter optimization
- Hands-on experience with infrastructure-as-code for AWS (e.g., Terraform) and automotive measurement data, such as MDF4 or MCAP
- Comfortable with relational databases (e.g., PostgreSQL) for metadata stores and experience with dataset management tools, functional-safety awareness (ISO 26262), or AUTOSAR Adaptive