Research Engineer in AI-driven Social Simulations (application deadline 31/07/2026)

Ensimag Alumni; Uppsala kommun; On-site; SEK 420,000 - 560,000; Part time

Job summary

Swedish University of Agricultural Sciences (SLU) Ecology Centre and Grimsö Research Station are seeking an ambitious research engineer to join a project building AI-driven large-scale simulations of environmental policy scenarios. You will develop multi-agent LLM-based infrastructure, manage a scalable simulation framework, and run Monte Carlo experiments on GPU HPC clusters.

The role focuses on integrating AI with ecology and political science to stress-test policies, with opportunities to

Qualifications

• A degree in computer science, engineering, physics, applied mathematics, or a related quantitative field (Master or engineering degree sufficient; PhD welcome but not required).
• Strong programming skills in Python and sufficient understanding of machine learning/deep learning to work effectively with modern generative AI systems.
• Experience with software engineering practices such as version control, testing, documentation, modular design, and reproducible workflows.
• Ability to work independently, debug complex systems under time pressure, and take ownership of technical infrastructure.

Responsibilities

• Build the LLM-agent infrastructure from an existing implementation, including multi-tier memory, affect state, and a strategic reasoning layer that tracks reputation, commitments, and mental models of other agents.
• Extend and maintain the multi-agent simulation framework that runs dozens of concurrent LLM instances communicating through structured message-passing protocols, and develop its validation methodology.
• Run large-scale Monte Carlo simulations on GPU-enabled HPC clusters (e.g., NAISS).
• Publish and disseminate project results, co-authoring scientific publications at the intersection of computational social science, ecology, security studies, and AI.

Skills

Python programming

Machine learning

Generative AI

Reproducible workflows

Independent work

Debugging complex systems

Education

Master's or engineering degree in a quantitative field

Tools

Version control

Testing

Documentation

Modular design

Job description

Postée le 14 juil.
Lieu : Grimsö or Uppsala, Sweden

• Contrat : CDD
• Rémunération : A négocier

Company: Swedish University of Agricultural Sciences (SLU).
The SLU Ecology Centre and Grimsö Research Station conduct research on sustainable agriculture and forestry, plant protection, nature conservation, and wildlife management, providing scientific knowledge to inform Sweden and Europe’s environmental policies.

Program overview

Can we foresee the outcomes of public policies, political choices and other decisions? We are finding out by developing serious games and running them using AI. We are building large‐scale simulations of societal processes (environmental negotiations, nature conservation policies, and hybrid‑threat scenarios) in which every actor is an autonomous AI agent powered by Large Language Models (LLMs). These agents simulate real‑world stakeholders, from government ministers to interest groups, and interact through natural language in complex strategic settings. We then run thousands of iterations to map the distribution of outcomes.

This effort is part of the new research program Articulating Complexity hosted at SLU and led by Guillaume Chapron (Docent). It is funded by grants from the Swedish Research Council (VR), the Swedish Foundation for Strategic Environmental Research (Mistra), and the Swedish Research Council for Sustainable Development (FORMAS). Our goal is to build a novel methodology at the intersection of AI, ecology, political science, and complex system analysis that may change how policies are designed and stress‑tested and how governments prepare for crises. We are looking for an ambitious research engineer to join this effort.

Tasks and duties

• Build the LLM‑agent infrastructure from an existing implementation, including multi‑tier memory, affect state, and a strategic reasoning layer that tracks reputation, commitments, and mental models of other agents.
• Extend and maintain the multi‑agent simulation framework that runs dozens of concurrent LLM instances communicating through structured message‑passing protocols, and develop its validation methodology.
• Run large‑scale Monte Carlo simulations on GPU‑enabled HPC clusters (e.g., NAISS).
• Publish and disseminate project results, co‑authoring scientific publications at the intersection of computational social science, ecology, security studies, and AI.

Required qualifications

• A degree in computer science, engineering, physics, applied mathematics, or a related quantitative field (Master or engineering degree sufficient; PhD welcome but not required).
• Strong programming skills in Python and sufficient understanding of machine learning/deep learning to work effectively with modern generative AI systems.
• Experience with software engineering practices such as version control, testing, documentation, modular design, and reproducible workflows.
• Ability to work independently, debug complex systems under time pressure, and take ownership of technical infrastructure.

Desirable qualifications

• Experience with LLMs, including programmatic use through APIs and local deployment; ability to select, configure, adapt, and critically evaluate LLMs for research applications.
• Experience with deep reinforcement learning, agent‑based modelling, or multi‑agent systems, including the design of simulation environments, interacting agents, decision rules, feedback mechanisms, and scenario‑based analyses.
• Experience with high‑performance computing: batch job submission, GPU workflows, and managing large‑scale computational experiments.
• Interest in nature conservation, environmental policy, political science, geopolitics and security studies, complex adaptive systems. A formal background in these fields is not required, but interest is essential.

Assessment criteria

• Demonstrated technical ability – portfolio, deployed system, or track record of solving hard AI problems will carry more weight than a list of credentials.
• Autonomy and resourcefulness – ability to diagnose problems independently, make pragmatic technical decisions, and keep complex systems running.
• Collaborative disposition – effective communication of technical constraints and possibilities to non‑technical collaborators.