Qualifications
Requirements:Demonstrated hands-on experience in the development and training of generative models, encompassing diffusion, flow matching, or masked-Transformer models. This ideally includes experience in training robotic imitation policies, generating whole-body motions for human avatars, or developing a vision-conditioned generative model to foresee actions. A PhD or master's degree in generative modeling, human motion generation, or learning-based robotic trajectory generation, complemented by relevant project experience. A robust research profile with a substantial record of publications in prestigious computer vision, graphics, and robotic conferences, such as ICCV/ECCV, CVPR, SIGGRAPH, CORL, RSS, etc. Proficient in Python and PyTorch, with practical experience in training and fine-tuning generative models, including diffusion, flow matching, and masked-Transformer models. Experience with the deployment of learning-based trajectory generation for robotic systems. Familiarity with modern GPU-based simulation environments like Omniverse or Genesis. Strong understanding of state-of-the-art machine learning architectures. Additionally, the following skills would be advantageous but are not mandatory:Experience with multi-modal generative models. Experience in image or video-based action prediction from ego-centric perspectives. Experience with ego-centric world models for human avatars or humanoid robots. Experience in fine-tuning foundation models, such as Gr00t or SmolVLA, to produce whole-body actions or kinematic motions. We seek an individual who thrives in a collaborative, fast-paced environment and is eager to take ownership of projects and decisions.
About the job
About Flexion Robotics:
At Flexion Robotics, we are pioneering the intelligence framework that will drive the next generation of humanoid robots. Our overarching mission is to expedite the transformation from delicate prototypes to their practical application in real-world scenarios. Founded by preeminent scientists specializing in robot reinforcement learning, with backgrounds from Nvidia and ETH Zürich, we are supported by top-tier international venture capital firms. In just a few months, we've transitioned from our initial line of code to the deployment of tangible humanoid capabilities alongside our partners.
The Role:
We are seeking a highly skilled expert in whole-body motion generation and/or vision-conditioned generative models to enhance our team in Zurich. Drawing inspiration from the recent advancements in neural avatars and computer graphics, we strive to implement similar methodologies to generate comprehensive body movements for our humanoid robots. The primary objective of this role is to innovate and implement cutting-edge multimodal generative models for robot trajectory generation.
About Flexion Robotics
Flexion Robotics is at the forefront of robotics innovation, creating intelligent systems that enhance the capabilities of humanoid robots. With a team led by experts in the field and strong backing from notable venture capitalists, we are transforming theoretical concepts into reality, pushing the boundaries of what humanoid robots can achieve.
