Connectionists: [jobs] MSCA 2026 Postdoc Fellowships in Open-Ended Embodied AI @ CYENS (Cyprus)

[Vassilis Vassiliades]

Dear all,

(Apologies for cross-posting)

The Cognitive Artificial Intelligence and Robotics (CAIR) team at the CYENS
Centre of Excellence (Cyprus) is inviting expressions of interest for
the *Marie
Skłodowska-Curie Postdoctoral Fellowships (MSCA-PF) 2026*.

We are recruiting ambitious candidates to co-develop competitive proposals
on *Open-Ended & Continual Learning, World Models, and Robot Learning*.

Please find the full call for expressions of interest below.

Best regards,
Vassilis Vassiliades

------------------------------

*[MSCA 2026 | Cyprus] Postdoctoral Fellowships: Open-Ended **Interactive &
Embodied Intelligence*

*Join the CAIR Team at CYENS to build generalist agents that discover,
reason, and act in interactive environments**.*



*The Opportunity*

The Cognitive Artificial Intelligence and Robotics (CAIR) research team at
CYENS Centre of Excellence (Nicosia, Cyprus) invites *expressions of
interest* for the *Marie Skłodowska-Curie Postdoctoral Fellowships (MSCA
PF) 2026.*

We are recruiting ambitious candidates to co-develop competitive proposals
on interactive and embodied AI. Our research vision moves beyond static
datasets toward generalist agents that leverage open-ended learning, world
models, and multi-agent dynamics to discover and adapt through interaction.
Projects may be evaluated in interactive environments (simulation, games,
and multi-agent worlds) and extended to real-world validation on robotic
platforms.
------------------------------

*Timeline*

   - *Call opens:* 9 April 2026
   - *Deadline:* 9 September 2026

------------------------------

*About the Host and Supervisor*

The CAIR team at CYENS is led by Dr. Vassilis Vassiliades and studies
learning systems that autonomously acquire, adapt, and reuse skills over
time. The supervisor’s prior research spans Quality-Diversity algorithms,
reinforcement learning, continual learning, and data-efficient robot
learning, with an interest in predictive models for planning and
decision-making.

These foundations motivate the team’s current direction toward open-ended
embodied intelligence. The MSCA project is intended as a collaborative
effort in which the postdoctoral researcher plays a central role in shaping
the scientific direction and establishing a new research line within the
team.
------------------------------

*Why Apply With CYENS?*

   - *The "Widening" Multiplier:* As a Cyprus-based host (a Widening
   country), eligible proposals may also be considered for ERA Fellowships if
   not funded under the main MSCA call, providing an *additional funding
   pathway* without requiring a new application.
   - *Pure Research Focus:* This is a dedicated research role with *no
   teaching obligations*, designed to support high-impact publications in
   venues such as NeurIPS, ICRA, CoRL, RSS, ICML, and ICLR.
   - *Agile Infrastructure:* We prioritize *compute-efficient* and
   *simulation-first* approaches, offering dedicated HPC resources (A6000
   48GB / A5000 24GB nodes) suitable for large-scale simulation and
   fine-tuning foundation models, with reliable access and minimal
   contention.
   - *International Secondments:* We actively facilitate optional research
   visits (3–6 months) to our network of international academic partners to
   strengthen training, networking, and the project's scientific impact.

------------------------------

*Research Tracks*

*We encourage proposals that bridge these areas (e.g., world models for
robust transfer and planning, or auto-curricula that drive the discovery of
reusable skills).* *Tracks 1–2 can be pursued fully in interactive
simulation environments, while Track 3 focuses on robotics and
cross-platform transfer.*

*1) **Open-Ended & Continual Learning*

*Moving beyond static datasets through self-generated experience and
lifelong interaction.*

   - *Auto-Curricula:* co-evolution of agents and environments (e.g., POET,
   UED) to generate "regret-maximizing" training distributions.
   - *Discovery Algorithms:* Quality Diversity (e.g., MAP-Elites),
   intrinsically motivated goal exploration, and unsupervised skill discovery.
   - *Continual & Meta-Learning:* solving the stability-plasticity dilemma
   and utilizing learning-to-learn strategies for fast adaptation to novel
   tasks without catastrophic forgetting.
   - *Predictive Knowledge:* learning large collections of self-generated
   predictions (e.g., General Value Functions) to continually expand the agent
   ’s knowledge without relying on external rewards.

*2) World Models & Abstraction*

*Structuring learned knowledge into abstractions that support planning and
reasoning** in embodied or interactive environments.*

   - *Multi-Level Representations:* hierarchical world models planning over
   skills/options rather than atomic actions.
   - *Learning Abstractions:* discovering compact state and skill
   representations that enable long-horizon planning.
   - *Multi-Agent **Models:* learning predictive models of other agents and
   joint dynamics for coordination and planning.
   - *System-2 Reasoning:* utilizing learned world models for test-time
   search (e.g., MCTS) and counterfactual reasoning.

*3) Robot Learning & **Transfer*

*Scaling robot learning via simulation-first development, portable data,
and cross-platform transfer (including sim-to-real when relevant).*

   - *Foundation Model Adaptation:* efficient fine-tuning/distillation of
   Vision-Language-Action (VLA) models for robotic control.
   - *Heterogeneous Coordination:* decentralized policies for diverse robot
   teams (e.g., Quadruped + Mobile Manipulator collaboration).
   - *Data-Driven Manipulation: *leveraging vision-based hand tracking,
   VR/AR interfaces, and handheld grippers for portable data collection,
   alongside low-cost arms for accessible teleoperation and imitation-learning
   pipelines.
   - *Robust **Transfer:* uncertainty-aware control and distributional RL
   for safe generalization across domains (simulation and real-world).

------------------------------

*Resources & Hardware*

We support simulation-first research with pathways to real-world validation
when relevant:

   - *Compute:* Dedicated CYENS AI Cluster (HPC) with A6000 48GB / A5000
   24GB nodes for large-scale simulation (Isaac Lab / MuJoCo) and deep
   learning.
   - *Robotics Platforms:* Unitree A1 (quadruped), Turtlebot 4 (mobile
   base), SO-101 low-cost manipulation arms for teleoperation and
   imitation-learning experiments; support for portable demonstration
   collection (UMI-style).
   - *Industrial Exposure:* Availability of commercial 6-DoF collaborative
   arms through CYENS facilities and partner arrangements.

------------------------------

*Eligibility Check*

   - *PhD Rule:* You must hold a PhD by *9 September 2026* (with max 8
   years post-PhD research experience).
   - *Mobility Rule:* You must *not* have lived or worked in *Cyprus* for
   more than 12 months in the 3 years prior to the deadline.
   - *Resubmission Rule:* If you applied for an MSCA-PF in the previous
   call (2025) with CYENS as your host and your proposal scored below 70%, you
   are not eligible to apply with us this year. (Applying with a different
   host previously does not affect your eligibility).

------------------------------

*How to Apply*

Email *v.vassiliades at cyens.org.cy <v.vassiliades at cyens.org.cy>* with the
subject: *“MSCA-PF-2026 Candidate”*

*Please include:*

   1. *CV:* Including a full list of publications.
   2. *Research Statement (Max 1 Page):* A concrete project sketch
   outlining the problem, method, and evaluation plan. *Please note:
   Generic applications without a specific research idea aligned with these
   tracks will not be considered.*
   3. *Representative Work:* PDFs or links to 2–3 of your best papers.
   4. *Eligibility Note:* Brief confirmation of your PhD date and months
   spent in Cyprus in the last 36 months, and confirmation that you do not
   fall under the MSCA 70% resubmission restriction with CYENS.


*Deadline for Expressions of Interest:* *15 May 2026* (Rolling review).

Shortlisted candidates will be contacted within 7–10 days of submission. We
expect to work closely with a small number of shortlisted candidates
(typically 2–3) to co-develop full MSCA proposals.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20260224/3cfbefbf/attachment.html>