Master Thesis: E2E AI model deployment and Closed loop Framework, ADAS Lab (m/w/d)"

About the position:

• The objective is to develop and validate a hardware-aware, closed-loop AI solution for urban automated driving. You will investigate how an existing end-to-end driving model can be efficiently deployed on a new automotive system-on-chip (SoC) and evaluate its behavior in realistic, near-real-time driving scenarios using a simulation environment.

Your team:

• Work in the TPP team within the ADAS Lab at Harman’s Garching office.

Your responsibilities:

• Model Deployment, Profiling, and Optimization
• Port an existing end-to-end driving model to an Automotive platform under Linux.
• Profile model execution at system, layer, and operator level.
• Analyze compute-engine allocation, memory behavior, bandwidth usage, latency, and performance bottlenecks.
• Investigate operator placement across CPUs and NPUs, custom operators, data orchestration, and other deployment optimizations.
• Explore hardware-aware model and operator designs suitable for future automotive SoCs.
• Derive practical recommendations for building accurate models that satisfy real-time execution and memory-bandwidth constraints.
• Reinforcement Learning and Closed-Loop Validation
• Review the current SIL and HIL-Framework for closed-loop simulation environment.
• Add reinforcement-learning train and evaluation methods to the SIL closed loop environment.
• Measure driving behavior, system performance, and end-to-end output latency.
• Generate internal simulation data.
• Define the reinforcement-learning algorithm, reward function, and training procedure together with the supervisors as part of the final thesis scope.

Your skills / qualifications:

• You are pursuing a master's degree in computer science, electrical engineering, robotics, automotive engineering, or a related field.
• You have a strong interest in automated driving, embedded AI, or machine learning deployment.
• You have practical experience with Python and a deep-learning framework such as PyTorch.
• You understand neural networks, model inference, and performance evaluation.
• Ideally, you have experience with reinforcement learning, Linux, automotive simulation, heterogeneous computing, or embedded accelerators such as NPUs.
• You work independently, approach technical problems systematically, and communicate your findings clearly.

What we offer: Flexible working time Training on the job from experts, and access to learning materials Above average payment Experiences and passion in a global acting team