Vision-Language-Action Models Intern

The AI Teams at Agile Robots are looking for an Intern (m/f/d) Vision-Language-Action Models, who will help train and evaluate the models that turn camera images and language instructions into robot movements. 

Responsibilities

• Experiments: Run training experiments and ablation studies, changing one part of the model, data, or training setup at a time and recording its effect on performance.
• Model Development: Implement and test changes to model code and training pipelines under the guidance of the researchers you work with.
• Robot Evaluation: Test trained models on real robots, measure how often they complete their tasks, and analyze the failures.
• Dataset Preparation: Prepare, clean, and analyze robot datasets that combine camera images, robot states, and actions, including checking that recordings are complete and aligned in time.
• Data Collection: Help collect new demonstration data on real robots, for example by teleoperating a robot arm through a task.

Essential Skills

• Education: Enrollment in a Master’s programme in computer science, robotics, machine learning, engineering, or a related field.
• Deep Learning: Hands-on experience training neural networks in Python with PyTorch or a comparable framework, for example in coursework, a thesis, or a project.
• Debugging: Writing maintainable code and tracing a failing run back to its cause in the data, the code, or the setup.
• Data Handling: Experience loading, cleaning, and analyzing large datasets in Python with libraries such as NumPy or pandas.

Beneficial Skills

• Robot Learning: Familiarity with vision-language-action models, imitation learning, or transformer-based robot policies.
• Robot Hardware: Hands-on work with real robots or robotics software, for example in a lab, a student team, or a thesis.
• Compute Clusters: Running and monitoring jobs on remote GPU machines or with a cluster scheduler such as SLURM.