PhD Research Internship - Robotics Engineer (VLM / VLA Models)

PhD Research Internship – Robotics Engineer (VLM / VLA Models) LocationBerlin / PotsdamEmployment TypeFull timeDepartmentEngineeringsensmore automates the world's largest machines with unprecedented intelligence. Our proprietary Physical AI enables heavy machines such as wheel loaders to instantly adapt to dynamic environments and execute new tasks without prior training.We integrate cutting-edge robotics into a platform powering intelligence and automation products - transforming productivity and safety for customers in mining, construction, and adjacent industries today.Join us and play a pivotal role in transforming the automation landscape in heavy industries.Role OverviewWe are seeking a highly motivated PhD candidate to join our team as a Research Intern specializing in General Purpose AI, with a focus on Vision-Language Models and Vision-Language-Action systems. This role sits at the frontier of industrial robotics: developing scalable, general-purpose VLA systems that enable robots to perceive, reason, and act autonomously in complex heavy-industry environments. You will contribute to bridging multi-modal perception (e.g., video, radar, lidar) with robust real-world execution, while advancing state-of-the-art methods in embodied AI. Beyond engineering, this position has a strong research component, with opportunities to contribute to novel methods, publish findings, and shape the future of industrial autonomy.Key ResponsibilitiesDepending on your expertise and project priorities, you will:Research & Method DevelopmentDesign and develop novel approaches for Vision-Language-Action systems in real-world industrial settingsExplore scalable architectures for multi-modal reasoning and action generationContribute to advancing state-of-the-art methods in embodied AI and robotic autonomyMulti-Modal Learning & Data SystemsLead the design and analysis of large-scale multi-modal datasets (video, radar, lidar, sensor fusion)Develop self-supervised or weakly supervised dataset generation pipelines for VLA trainingInvestigate data-centric approaches to improve robustness and generalizationModel Development & OptimizationBuild, adapt, and extend cutting-edge GenAI models (e.g., VLMs, VLA frameworks)Apply advanced fine-tuning strategies (e.g., parameter-efficient tuning, alignment methods)Explore prompt optimization, reasoning augmentation, and action grounding techniquesTraining, Evaluation & BenchmarkingDesign rigorous evaluation protocols for embodied AI systems in industrial contextsRun large-scale experiments, analyze performance, and iterate systematicallyBenchmark models against state-of-the-art approaches and internal baselinesDeployment & Systems IntegrationCollaborate with engineering teams to transition research prototypes into production-ready systemsOptimize models for real-time inference, robustness, and safety in heavy-industry environmentsScientific ContributionDocument findings and contribute to research publications, technical reports, or patentsPresent results internally and potentially at leading conferencesRequired QualificationsCurrent enrollment in a PhD program in Robotics, Computer Science, Machine Learning, Electrical Engineering, or a related fieldStrong programming skills in Python and deep learning frameworks (e.g., PyTorch)Solid understanding of machine learning, deep learning, and multi-modal modelsProven ability to conduct independent research and drive projects from idea to resultsStrong analytical thinking and problem-solving skillsPreferred Skills & ExperienceExperience with Vision-Language Models, embodied AI, or robotics learning systemsFamiliarity with modern GenAI tooling (e.g., Hugging Face ecosystem, Gemini, Unsloth, or similar)Experience with multi-modal data (vision + sensor fusion)Background in robotics, control systems, or real-world deploymentTrack record of research output (publications, preprints, or significant research projects)Experience with large-scale training, distributed systems,...