F

VLA/WAM Research Engineer

Flexion Robotics·Zürich, Zurich, Switzerland·03.11.2026

 jobs.workable.com

 
 
Festanstellung80–100%
Job written in
English
Location
Zürich, Zurich, Switzerland
Work type
On-site

 

Flexion is a fast‑growing robotics company creating the intelligence layer for next‑generation humanoid robots. Backed by top venture capital and founded by former Nvidia and ETH Zürich researchers, the firm is moving from prototype to real‑world deployment of humanoid capabilities. The opening is for a VLA/WAM Research Engineer who will build the foundation models that drive these robots, covering everything from model architecture to large‑scale pre‑training, adaptation, and continual improvement. The day‑to‑day work revolves around designing multimodal models that fuse vision, language, robot state, actions and temporal context. You will lead large‑scale pre‑training on diverse embodied datasets, including teleoperation, egocentric human video and simulated experience, while refining data mixtures, objectives and optimization strategies. The role also includes creating predictive world‑action models for representation learning and planning, devising efficient adaptation methods for new tasks or embodiments, and applying post‑training techniques such as imitation learning, reinforcement learning and specialist‑to‑generalist distillation. Continuous learning from autonomous rollouts, failures and interventions, as well as rigorous experimental design and evaluation on both simulation and real robots, are key parts of the job. Candidates must have strong Python and PyTorch expertise and a proven track record in research or engineering on at least one of the following: VLA/VLM or multimodal model architecture, large‑scale foundation‑model pre‑training, world models, robot learning or reinforcement learning/post‑training. You should have owned substantial model development or training experiments, making decisions on architecture, objectives, optimization and evaluation. A solid grasp of modern deep‑learning architectures, representation learning and the link between data, objectives and performance is required, together with the ability to design rigorous experiments, diagnose training failures and translate empirical results into research direction. Familiarity with large‑scale datasets and model training workflows, plus the skill to critically assess recent research and turn promising ideas into working prototypes, are also essential. Nice‑to‑have experience includes training VLA, VLM, diffusion, flow‑matching or video/world generative models, as well as working with distributed multi‑node GPU systems and large‑scale data mixtures. Background in reinforcement learning (offline RL, reward/value learning), specialist distillation, continual learning, predictive world models, latent dynamics or self‑supervised representation learning is valued. Experience with multi‑embodiment robot learning, action representations, policy adaptation across platforms, and handling teleoperation, egocentric video, simulation or autonomous rollout datasets is also a plus. Direct evaluation of learned policies on real humanoid or dexterous manipulation robots would be advantageous. The position is based in Flexion's central Zürich office, which houses state‑of‑the‑art robotics testing facilities. The company offers competitive compensation, an enhanced pension plan, generous holiday and paid‑leave perks, relocation and permit sponsorship, as well as health, dental and vision coverage with flexible plan choices. The culture is energetic and collaborative, emphasizing bias for action, regular community events and exposure to groundbreaking research within Europe's leading robotics team. What the role asks for: - Strong Python proficiency - Strong PyTorch proficiency - Research or engineering experience in VLA/VLM architectures - Research or engineering experience in multimodal foundation models - Research or engineering experience in world models or predictive learning - Research or engineering experience in large‑scale pre‑training - Research or engineering experience in robot learning or RL/post‑training - Experience owning substantial model development or training experiments - Nice-to-have: Experience training VLA, VLM, diffusion or video/world models - Nice-to-have: Experience with distributed multi‑node GPU training - Nice-to-have: Experience in reinforcement learning, offline RL or reward/value learning - Nice-to-have: Experience in continual learning or specialist‑to‑generalist distillation - Nice-to-have: Experience with multi‑embodiment robot learning and policy adaptation.

 

 

 

 

 

Kostenlos

Engineer: Lebenslauf-Vorlage

Bewirb dich mit einem Lebenslauf im Schweizer Aufbau — mit Beispieltext und den Anforderungen, die Engineer-Inserate am häufigsten nennen.

Lebenslauf-Vorlage Schweiz ansehen

Engineer: was der Markt gerade verlangt

Engineer-Stellen gehören zu den regelmässig ausgeschriebenen Berufen auf SwissJobs.app.

Weitersuchen

Ähnliche Jobs per E-Mail erhalten