Machine Learning Engineer / Edge-ML Automotive (m/w/d/kA)

We are the KOSTAL Group, a globally operating, independent family business headquartered in Germany, developing and producing technologically sophisticated electronic and mechatronic products. More than 20,000 employees work flexibly, competently, and with a customer-centric approach at 64 locations in 22 countries. The KOSTAL Automotive Electrical division develops and produces power electronics for electromobility, control units for comfort electronics, and user interfaces ranging from turn signal switches to touchscreens.

• Development of robust ML algorithms for intent/gesture recognition based on vehicle-side sensor data (e.g., UWB radar), including real-time inference on edge ECUs.
• End-to-end ownership of the ML pipeline: data acquisition on the prototype vehicle, ground truth/labeling, feature engineering, model training (PyTorch), experiment tracking (MLflow), model versioning and ensuring reproducibility
• Edge deployment: Quantization/optimization (e.g., TorchScript/ONNX) to adhere to strict latency and resource budgets under real-world conditions.
• MLOps setup and operation: CI/CD (e.g., Jenkins), containerization (Docker/Dev Containers), test automation (unit, integration, and on-car validation), model monitoring and drift
• Definition of data and test campaigns, error analyses and robustness considerations
• Creation of technical documentation (model and data maps, evaluation reports, KPI tracking) and presentation of results

• Successfully completed studies in computer science, electrical engineering, mathematics, physics or a comparable field.
• Several years of professional experience in applied ML/Data Science, preferably in the embedded/automotive environment.
• Excellent knowledge of Python and PyTorch; proficient in using MLflow for experiment tracking and model management.
• Practical experience in sensor data processing and fusion (e.g., UWB, radar, IMU; optionally vision), signal processing, tracking (e.g., Kalman/particle filters), and robust classification/detection methods.
• Experience with edge optimization: quantization, pruning, TorchScript/ONNX; performance and latency tuning under resource constraints
• Proficient in MLOps: Docker/Dev Containers (VS Code), Git/GitLab CI, reproducible training pipelines, data/model versioning (e.g., DVC), test automation
• Knowledge of embedded Linux/C++ is advantageous; experience with on-car data acquisition (e.g., CAN/FlexRay, ROS2) and experimentation is a plus.
• Ideally, experience in the automotive environment (pre-development/near-series development)
• Very good German and good English skills

• Values-oriented work culture - We are down-to-earth, appreciative, inspiring and innovative.
• Attractive contract terms - at least 30 vacation days and above-standard pay for performance that pays off
• High flexibility - trust-based working hours with plenty of personal freedom and options for mobile working.
• Personal development - Individual onboarding, extensive training opportunities and career prospects for long-term collaboration, also internationally
• Exciting extras - modern company restaurants, company pension scheme, bike leasing and numerous corporate benefits
• Sustainability in practice - responsible use of resources and development of efficient solutions for drive technology, e-mobility, charging technology and photovoltaics

For further information about this job offer, please contact André Kohlschütter at +49 171 5539 445. We look forward to your interest and your application.