Applied Research Scientist in Cognitive Robotics & Robot Reasoning

Are you excited about building the next generation of robots that can understand situations, reason about goals, deliberate under uncertainty, and act in the physical world?

The Cognitive Robotics Group within the Robotics and Automation Unit at Eurecat Technology Centre is looking for an Applied Research Scientist in Cognitive Robotics and Robot Reasoning.

You will work on GRAIL (Grant Agreement No. 101298421) - a flagship European project developing foundation models for robotics - and continue contributing to Eurecat's cognitive robotics roadmap beyond the project. This is a permanent research position.

We are looking for someone with PhD-level research maturity - either through a completed PhD or equivalent research experience - and hands-on experience applying AI/ML, reasoning, planning, or foundation-model methods to robotic systems.

Why this role matters Robotics is entering a new phase: generative AI and foundation models are changing how robots perceive, interpret, decide, and acquire new capabilities. But robotics is not only language. Robots need grounded reasoning, spatial and semantic world understanding, memory, uncertainty handling, safety, and a reliable connection between deliberation and physical action.

In this role, you will help bridge that gap: turning advanced AI methods into cognitive robotic capabilities for task reasoning, situational grounding, world modelling, and decision-making in real industrial and semi-structured environments.

This is a unique opportunity to work with some of the strongest robotics and AI groups in Europe on a project that aims to shape the future of physical AI and industrial automation. If you want your research to move beyond papers and benchmarks into real robotic systems that reason, explain, adapt, and act, this is the kind of opportunity you should not miss.

What You Will Work On You will contribute to applied research and development in areas such as:

Robot reasoning, deliberation, task planning, and decision-making for embodied agents operating under uncertainty.
Foundation models for robotics, including multimodal models that connect language, vision, knowledge, memory, scene understanding, and action-relevant representations.
Design and implementation of new robotics-oriented AI architectures and agentic systems, combining foundation models with planners, memory, tool use, world models, verifiers, simulators, and robot-action interfaces.
Grounding of reasoning in the physical world: object and scene understanding, affordances, spatial/semantic context, task constraints, human instructions, and feedback from real robot execution.
Predictive world models, simulation-supported reasoning, data-driven adaptation, and methods that help robots generalise across tasks, environments, objects, and platforms.
Scientific publications, project deliverables, demonstrations, and collaboration with European partners from robotics, AI, industry, and human-centred technology.
You will not be working only on isolated academic benchmarks or language-only prototypes. The role involves developing ideas, implementing them, testing them with robotic systems and realistic tasks, and adapting them to practical constraints.

Requirements:

What We Are Looking For We do not expect you to match every keyword. We are looking for depth, curiosity, and the ability to turn research ideas into working cognitive robotic systems.

Essential
PhD in Robotics, Artificial Intelligence, Machine Learning, Computer Vision, Cognitive Systems, Automated Planning, or a related field; or equivalent research experience.
Strong background in machine learning, deep learning, symbolic/neural reasoning, planning, or foundation models applied to robotics or embodied AI.
Hands-on experience designing, training, fine-tuning, or adapting AI model architectures for robotics or embodied AI, including implementation choices around data, losses, evaluation, and deployment - not only API use or prompt engineering.
Experience connecting perception, language, knowledge representations, planning, decision-making, and action in embodied or robotic systems.
Strong programming skills in Python.
Ability to work with real robotic systems, experimental platforms, or realistic robotic demonstrators - not only offline datasets or simulation.
Scientific and technical maturity: ability to define problems, test hypotheses, analyse results, integrate components, and communicate findings clearly.
Good written and spoken English.
Highly valuable
Experience with LLMs, VLMs, multimodal foundation models, transformers, neuro-symbolic methods, task and motion planning, hierarchical planning, memory architectures, or uncertainty-aware reasoning.
Experience with end-to-end model development and training pipelines for robotics foundation models or agentic systems: multimodal data curation, pre-training or fine-tuning, PEFT/LoRA, synthetic or simulation data, ablations, benchmarking, and real-robot validation.
Experience with ROS/ROS 2.
Experience with PyTorch, TensorFlow, JAX, or similar frameworks.
Experience with semantic or 3D scene understanding, affordance modelling, world models, simulation, manipulation-aware reasoning, human-robot interaction, or human instruction following.
Publications, open-source contributions, project demos, or research prototypes in cognitive robotics, robot reasoning, embodied AI, foundation models for robotics, or robot learning.
Experience in European or collaborative R&D projects.
This role is probably a good fit if...
You have worked on AI methods that help robots understand a situation, reason about what to do, explain or justify decisions, adapt to new contexts, or connect high-level goals to physical action.
You enjoy both research and implementation.
You are comfortable moving between papers, code, experiments, robots, datasets, simulations, and project discussions.
You want to work in applied research: scientifically ambitious, but connected to real systems, demonstrators, and industrial impact.
You are attracted by the opportunity to contribute to a European flagship effort in robotics foundation models, with a strong focus on reasoning, grounding, and trustworthy embodied intelligence.
This role is probably not the best fit if your main experience is limited to...
Generic LLM chatbot development without robotics, embodiment, planning, or grounding in the physical world.
Language-only evaluation or prompt engineering with no connection to robot perception, action, or decision-making.
Pure navigation, PLC programming, or classical control with no AI, reasoning, or robot-learning component.
Simulation-only research with no interest in deploying, validating, or adapting methods on real robotic systems.
Software front-end or general AI engineering unrelated to physical robotic systems.
What Eurecat Offers Eurecat is a leading technology centre in Catalonia, working at the intersection of research, technology transfer, and industrial innovation.

In our Robotics and Automation Unit, the Cognitive Robotics Group develops applied research in robot cognition, reasoning, behaviour, perception, human-robot interaction, autonomous systems, manipulation-aware intelligence, and foundation models for robotics.

You will join a team with strong experience in European projects, industrial collaborations, and real robotic demonstrators. Our laboratory is located in Cerdanyola del Valles, near Barcelona, within a growing robotics and AI ecosystem.

Additionally, you will have these benefits:

Permanent contract.
Hybrid work.
Flexible schedule.
Shorter workday on Fridays and summer schedule.
Flexible compensation package: health insurance, transport, lunch vouchers, training, kindergarten, and other benefits.
Access to Eurecat Academy courses.
Language training in English, Catalan, and Spanish.
Participation in a major European robotics and AI project with leading academic and industrial partners.
The opportunity to build long-term research lines in cognitive robotics and robot reasoning beyond a single project.