Lead Data Engineer

Germany | Full-time, on-site | Competitive six-figure base We're working with a fast-growing deep tech company in Germany that is developing AI for machines operating in the physical world. As their models scale, so does the volume and complexity of their real-world data, and they're hiring a senior engineer to take ownership of how that data is collected, processed and turned into better models. This is a hands-on, high-impact role sitting between software and machine learning, with genuine technical authority and a direct line to senior leadership. The role Design and run large-scale distributed processing for video, sensor and time-series dataBuild the infrastructure that gets data efficiently from collection into large GPU training runsWork closely with ML researchers to understand where models fall short, and fix it through better dataDefine how multimodal data is structured, stored, versioned and quality-checkedDevelop Python tooling for validation, cleaning and dataset managementHelp shape data collection, annotation processes, and secure handling of sensitive data What you'll brin Significant experience in data engineering, ideally with large volumes of real-world data from robotics, autonomous driving, drones or video AIStrong hands-on experience with distributed processing frameworks such as Ray, Spark, Dask or BeamExpert, production-grade Python, including asyncio, multiprocessing and performance optimisationStrong working knowledge of PyTorch and NumPy, and of how training workloads consume dataA proven track record building pipelines that feed multi-node GPU training on AWS or GCPExperience with high-throughput storage and data loading for training (e.g. S3/GCS, Parquet/Arrow, WebDataset or similar)Hands-on experience with video encoding (H.264/H.265, FFmpeg) and robotics or sensor log formats (e.g. ROS bags, MCAP or similar)Dataset versioning and lineage at scale using tools such as DVC, LakeFS, Delta Lake or IcebergA solid grasp of machine learning fundamentals, enough to judge which data actually improves a model Nice to have Building training datasets for multimodal or foundation models (e.g. VLA, VLM)Time synchronisation across high-frequency sensor streams and 3D coordinate transformsActive learning, data-centric ML or failure-case mining workflowsWorkflow orchestration (Airflow, Dagster, Prefect) and containerised infrastructure (Docker, Kubernetes)Experience working with annotation vendors or data collection operations Why it's worth a conversation A senior hands-on role with real ownership and influenceWork on cutting-edge AI with immediate, visible impactA strong engineering culture in a company with serious backing #J-18808-Ljbffr