Requirements
- PhD in Computer Science, a related field, or equivalent practical experience.
- Experience with vision, vision-language, video, and other multimodal models, especially ones trained with human data.
- Experience with multimodal generative modeling, training, and inference.
- Experience with reinforcement learning and imitation learning.
Preferred qualifications:
- Experience with capture methodologies, dataset design, experimentation, and incorporation of captured human action data into VLA, WAM, or RL training.
- Experience with simulators and real-world robots, especially dexterous manipulation, as well as multimodal sensing (e.g., tactile, forces).
About Flexion
At Flexion, we're building the intelligence layer powering the next generation of humanoid robots. Our mission is to accelerate the transition from fragile prototypes to real-world humanoid deployment. We are founded by leading scientists in robot reinforcement learning (ex-Nvidia, ex-ETH Zürich), and backed by leading international VC firms. In just months, we’ve gone from our first line of code to deploying real humanoid capabilities.
The role
We are seeking an expert in robotics foundation models and multi-modal human action data to advance the frontiers of generalizable manipulation. This position bridges the gap between human embodiment and robotic execution by leveraging video demonstrations, wearable capture systems (such as UMI or sensorized gloves), and complementary data sources. You bring a deep understanding of the full lifecycle required to scale dexterous capabilities, spanning the design and deployment of specialized capture devices, dataset curation, and algorithmic and architectural modeling.
As a Research Scientist/Engineer, you will drive large-scale experimentation and deploy breakthrough ideas across core physical intelligence problems. You will own the development of novel systems that turn cutting-edge research into robust, real-world implementations, ultimately pushing the limits of what our hardware can achieve.
Key responsibilities
- Scalable Human Data Integration: Develop methodologies and capture pipelines to integrate human video demonstrations, wearable systems (such as UMI or sensorized gloves), and unconstrained manipulation data directly into foundation model and RL workflows.
- 3D Object & Interaction Reconstruction: Build automated pipelines to reconstruct manipulated 3D objects and environments from capture data, importing digital twin assets into simulation to enable scalable scenario synthesis.
- Simulation-to-Real & Reinforcement Learning: Build simulation environments with reconstructed assets, leveraging reinforcement learning, tactile, and force feedback to produce dexterous manipulation policies.
- Multimodal Policy Architecture & Evaluation: Design, train, and evaluate robot foundation models that translate multi-sensor data streams into robust control.
- Embodied Transfer: Design algorithms to bridge human embodiment to high-DOF dexterous hands and humanoid platforms.
- Competitive compensation
- Enhanced pension plan
- Enhanced holiday & paid leave perks
- Relocation & permit sponsorship
- Central Zürich office with top-tier robotics testing facilities and infrastructure
- Joining Europe's leading robotics team & exposure to never-done-before research
- Energetic, collaborative culture with a bias for action and regular community events
- Competitive Compensation
- Joining a leading robotics team & exposure to never-done-before research
- Energetic, collaborative culture with a bias for action and regular community events
Zurich
- Enhanced pension plan
- Relocation & permit sponsorship
- Enhanced holiday & paid leave perks
- Central Zürich office with top-tier robotics testing facilities and infrastructure
San Franciso
- 401(k) with company contributions
- Health, dental & vision coverage with the flexibility to choose your own plan
- Open PTO policy & paid company holidays
