Research Scientist: Robot Foundation Models & Dexterity

Enact; Zürich; On-site; CHF 150,000 - 190,000; Full time

Benefits offered by this job

Visa sponsorship

Relocation support

Equity participation

Central Zürich office

Job summary

Enact is building the base intelligence layer for robotics, training large-scale force-aware foundation models to unlock dextrous manipulation across tasks and environments. You will design and run large-scale pre-training runs for robot foundation models, defining architectures and curricula to push AI-driven manipulation beyond pixels to actions.

You will own experiments end-to-end, from data specification through training to real-robot evaluation, to understand scaling laws, data quality

Qualifications

• PhD or equivalent research experience in ML, CV, robotics or related fields.
• Experience designing rigorous experiments and separating signal from noise.
• Strong fundamentals in ML computing frameworks and scalable training.

Responsibilities

• Design and run large-scale pre-training for robot foundation models and define architectures and training curricula.
• Own experiments end-to-end from data spec to real-robot evaluation, studying scaling laws and data quality effects.
• Build data mixes and evaluation sets to train and assess robots on novel high-precision tasks across real-world and simulated settings.
• Research methods to improve sample efficiency and robustness for deployment in real-world robotics.

Skills

Large-scale models

Experiment design

ML fundamentals

Distributed training

Prototyping robotics

Education

PhD or equivalent research experience

Tools

PyTorch

JAX

Job description

Enact is building the base intelligence layer for robotics, training large-scale force-aware foundation models to unlock dextrous manipulation across tasks and environments. You will design and run large-scale pre-training runs for robot foundation models, defining architectures and curricula to push AI-driven manipulation beyond pixels to actions.

You will own experiments end-to-end, from data specification through training to real-robot evaluation, to understand scaling laws, data quality