AI Engineer – Human-to-Robot Learning (human)

Neura Robotics; Riederich; Vor Ort; EUR 90.000 - 130.000; Vollzeit

Zusammenfassung

Neura Robotics is seeking a research-oriented professional to advance multimodal foundation models for robotic systems in Germany. The role centers on turning sensor data into learnable representations and translating human motion across bodies for operative robot control.

The candidate will explore cross-embodiment transfer, imitation and self-supervised learning, while collaborating with hardware and systems teams to ensure robust AI solutions.

Qualifikationen

• Extensive hands-on work with multimodal foundation models.
• Experience with cross-embodiment transfer and RL-based retargeting is required.
• Deep understanding of robotics control at a low level.
• Strong grounding in imitation, representation and self-supervised learning across sensor modalities.
• Fluency with modern ML frameworks and large-scale training infrastructure.
• Effective communication across researchers, engineers and hardware teams.

Aufgaben

• Model the Human Experience by converting multimodal data into learnable representations for robots.
• Bridge two bodies by translating human motion/interactions into usable robot action spaces.
• Design pretraining and fine-tuning strategies for robotic foundation models.
• Prove it works by building benchmarks that distinguish generalized models from mere imitators.
• Shape data-capturing requirements by identifying model blind spots.
• Collaborate across hardware and systems engineering while focusing on data and AI tasks.

Kenntnisse

Multimodal models

Cross-embodiment transfer

Imitation learning

Self-supervised learning

Large-scale training infra

English fluency

German language (plus)

Ausbildung

Master’s or PhD in CS/ML/Robotics

Jobbeschreibung

Your Mission & Challenges

• Model the Human Experience: Turn multimodal, body-worn data into representations a robot can learn from.

• Bridge Two Bodies: Solve cross-embodiment transfer, translating human motion and interaction into action spaces a robot with a completely different body can use.

• Build Foundation Models That Act: Design pretraining and fine-tuning strategies for robotic foundation models.

• Prove It Works: Build the benchmarks that separate models that merely mimic from models that generalize.

• Shape What Gets Captured Next: Turn model blind spots into sharp, concrete requirements for data capturing devices.

• Cross-Functional Collaboration: Work shoulder to shoulder with hardware and systems engineering teams while staying laser-focused on the data and AI side.

What We Can Look Forward To

• Master’s or PhD in Computer Science, Machine Learning, Robotics, or comparable.

• Real, hands-on experience with multimodal foundation models (VLA, video-action, or similar).

• Experience with cross-embodiment transfer, RL based retargeting.

• Deep understanding of low-level robotics control.

• Strong grounding in imitation learning, representation learning, and self-supervised learning across sensor modalities.

• Fluency in modern machine learning frameworks and large-scale training infrastructure.

• A communicator who moves easily between researchers, engineers, and hardware teams.

• Fluent English, German a plus.