Forward Deployed AI Engineer

Agile Robots SE is an innovative company with a clear mission: bringing intelligent robotic solutions to our customers. We combine the latest technology in the field of AI with powerful robotic systems — from industrial robot arms and mobile manipulators to humanoids — to solve tasks that classical automation cannot address.

At the heart of our offering is a versatile software platform for orchestrating complex robotic systems. It allows efficient customization for specific customer requirements: configuring systems, developing reusable skills, parameterizing processes, and executing tasks reliably in production. Learned models — from perception to end-to-end policies — are a core part of this platform.

Our solutions are deployed with international customers across a wide range of industries. Seamless integration and scalable deployment are at the core of our mission.

We are looking for a passionate Forward Deployed AI Engineer to bring state-of-the-art robot learning and computer vision models into real production environments. In this role, you will train, fine-tune, and deploy AI models directly on customer robots — from Vision-Language-Action models and diffusion policies to perception and grasp generation networks. You will become a trusted technical partner, turning customer challenges into robust, data-driven robotics applications. The role is based in Munich, Germany.

Responsibilities

• Deploy, fine-tune, and evaluate AI models on real robotic systems at customer sites — including Vision-Language-Action (VLA) models, diffusion policies, and other end-to-end learned policies
• Adapt and train perception models for robotic applications, e.g. object detection, segmentation, and pose estimation based on YOLO, SAM, vision transformers, and self-supervised backbones such as DINOv2
• Integrate and tune grasp generation models (e.g., GraspNet, Dex-Net, Contact-GraspNet) for bin picking and manipulation tasks
• Build data pipelines for customer deployments: demonstration and teleoperation recording, dataset curation, annotation, training runs, and systematic evaluation of model performance
• Bring models into production: optimize inference, deploy on edge and embedded hardware, and integrate models into our robotic software platform
• Work closely with customers to understand technical requirements and translate them into robust, AI-driven robotics applications
• Diagnose, troubleshoot, and resolve model, software, and system integration issues both remotely and on-site
• Travel to international customer sites for workshops, deployments, training, and technical support (up to 40% travel)

Essential Skills

• Master's degree in Robotics, Computer Science, Artificial Intelligence, Machine Learning, or a related field — or equivalent practical experience
• Minimum of four years of professional experience in machine learning, AI for robotics, or computer vision
• Hands-on experience working with real robots — robot arms, mobile manipulators, grippers, cameras, and sensors — not only in simulation
• Proven experience with end-to-end learned models for robotics (e.g., VLAs, diffusion policies, imitation or reinforcement learning policies), covering training, fine-tuning, and deployment
• Strong programming skills in both Python and C++
• Solid command of machine learning frameworks (e.g., PyTorch, TensorFlow) and the full model lifecycle: data curation, training, evaluation, and deployment
• Experience with modern computer vision models such as YOLO, SAM, vision transformers, and self-supervised backbones (e.g., DINOv2, CLIP)
• Experience with learned grasp generation for manipulation (e.g., GraspNet, Dex-Net, or comparable approaches)
• Good understanding of robot (inverse) kinematics, coordinate transformations, and hand-eye calibration
• Experience working in Linux environments
• Willingness to learn new technologies and adapt quickly in customer environments
• Fluency in English

Beneficial Skills

• Experience with large-scale demonstration data collection, teleoperation setups, and sim-to-real transfer
• Experience with robot learning simulation frameworks (e.g., NVIDIA Isaac Sim/Lab, MuJoCo, Gazebo)
• Good understanding of classical computer vision (e.g., OpenCV, PCL, camera calibration, point cloud processing)
• Experience optimizing inference (e.g., TensorRT, ONNX, quantization) and deploying on NVIDIA embedded platforms (e.g., Jetson Orin, AGX, Xavier)
• Experience with MLOps tooling: experiment tracking, dataset versioning, and continuous training or evaluation pipelines
• Publications or open-source contributions in robot learning, manipulation, or computer vision
• Familiarity with industrial automation environments (e.g., PLCs, EtherCAT, OPC UA)