Postdoctoral Researcher in Multi-Agent LLM Systems for the Edge–Cloud Continuum

• Position: Postdoctoral Researcher
• Contract: Temporary Position Longer Than 6 Months
• Employment: Full-Time
• Location: Ljubljana, Slovenia (On-Site)
• Salary: € 2900 – 4300 monthly gross (depending on qualifications)

We are hiring a postdoctoral researcher to help lead a research line on effective collaboration in multi-agent LLM systems, brought to real edge–cloud infrastructure.

The opportunity

We are looking for a postdoctoral researcher to help lead the following research line: understanding and engineering effective collaboration in multi-agent LLM systems, and bringing it to real edge–cloud infrastructure.

Our recent work shows that adding more agents to an LLM system yields sharply diminishing returns — a debate of 30 agents can collapse to an effective team size of fewer than two as their answers converge. We frame this through the Ringelmann effect and quantify it with information-theoretic measures of the independent perspectives a system actually contributes.

Some of the EU projects we work on take this question to the edge–cloud continuum, where compute, energy, latency, and bandwidth are all constrained and heterogeneous. This is a natural fit: heterogeneous hardware invites heterogeneous models, network structure shapes agent communication topology, and energy budgets make efficiency a first-class objective. You will help turn “effective team size” into a design metric that sits alongside latency and cost.

What you’ll work on

• Scaling laws for effective collaboration — measuring and maximizing the independent perspectives (effective team size, error decorrelation) in multi-agent LLM systems, and characterizing where the benefit of scale runs out.
• Heterogeneous model teams matched to heterogeneous compute — combining small on-edge models with larger cloud models so their errors decorrelate rather than compound.
• Communication topologies and orchestration under real constraints — how the graph of who-talks-to-whom, and where agents run across edge and cloud, drives (or resists) convergence, under latency, bandwidth, and energy limits.
• Energy- and cost-aware agent orchestration — efficiency-first placement and aggregation strategies across the continuum.
• Task-appropriate aggregation — matching combination rules (vote, median, pool, judge) to the nature of the task.

You’ll publish at top ML/systems venues, contribute to our open-source stack, and shape the direction of the group’s agenda in this area.

The role — research and leadership

This is a mixed state and EU-project position, so beyond research excellence you will grow into a leadership role:

• Project management — coordinating tasks, deliverables, and milestones with European partners, and representing the group in project meetings.
• Proposal writing — contributing to (and in time leading sections of) follow-on funding proposals.
• Student mentoring — co-supervising and guiding MSc/PhD students and junior researchers day to day.

We are open to candidates who have these skills or who are genuinely eager to develop them — we will actively support that growth.

Your profile

Required

• PhD (obtained or near completion) in Computer Science, Computer Engineering, Electrical Engineering, or a related field.
• Strong background in one or more of: LLMs / multi-agent systems, distributed / edge–cloud computing, ML systems.
• Solid experimental and software skills (e.g. Python; experience running experiments on GPU/HPC clusters).
• A track record of peer-reviewed publications appropriate to career stage.
• Excellent written and spoken English.

Desirable

• Experience with multi-agent orchestration frameworks, model serving (e.g. vLLM), or edge inference.
• Familiarity with EU-project ways of working (deliverables, partner coordination) — or clear motivation to learn.
• Experience mentoring students and/or contributing to grant proposals.
• An interdisciplinary, curious mindset that connects ideas from group psychology, information theory, and distributed systems.

What we offer

• A visible, high-impact research topic at the frontier of agentic AI, with room to define your own direction.
• A collaborative, international team and a clear path toward research independence (proposal writing, supervision, project leadership).
• Competitive salary per national/EU scale: € 2900 – 4300 monthly gross depending on qualifications. Additionally, monthly lunch fees (€ 150) and local transportation to work and back (bus, etc.) are provided. A total of € 1500 holiday allowance is also provided as per national laws. Funding for travel to conferences and open access journal fees is available.
• One year full-time contract, with possibility of extension and a clear path to tenure.
• Ljubljana, Slovenia — a quality of life index of 178.40; estimated cost of living available at Numbeo.

How to apply

Please send your CV, publication list, a short (≤1 page) statement of research interests, and contact details for 2 referees to sensorlab@ijs.si with subject “Postdoc — [your name]”. Applications are reviewed on a rolling basis until the position is filled. Informal enquiries to carolina.fortuna@ijs.si.

Jožef Stefan Institute is an equal-opportunity employer and welcomes applications from all qualified candidates regardless of background.