Machine Learning Engineer (m/f/d) VLA Training & Evaluation
Posted Updated
Description
Agile Robots SE, Munich-based, seeks researchers to advance training, fine-tuning, and evaluation of vision-language-action models for robotic control. You will design experiments, create reusable pipelines, and establish baselines and templates for others to extend.
The role emphasizes transfer studies between simulation and real hardware, with opportunities for impact across cross-disciplinary teams and real-world robotics projects.
Training workflows: build reusable, model-agnostic training and evaluation workflows with experiment tracking and model versioning. Fine-tuning pretrained vision-language-action models on robotics data using LoRA/adapters or distillation. Policy training loops for action-chunking and transformer-based models, diffusion policies, and VLM/VLA heads. Benchmarks with generalization splits and failure-analysis templates. Transfer studies with simulation engineers to assess sim-to-real improvements. Research baselines turning published methods into working baselines and templates. Academic background in Computer Science, ML, Robotics or related field (Masters/PhD typical) or equivalent experience. Hands-on experience writing and debugging training loops in PyTorch or JAX, including data loading, loss design, and optimizer behavior. Experience fine-tuning pretrained models and adapting them to new tasks. Understanding imitation-learned policies and how they turn observations into robot actions. Experience designing evaluations and ablations with tracking, versioning, and checkpoint management. Corporate Benefits Program covering 健康 Modern Munich office with rooftop Regular company eventsSkills
Sign in to read and join the discussion.