My research investigates the mechanisms that shape the behavior of LLM agents. More broadly, I am interested in how AI models influence human learning, reasoning, creativity, and social organization. I also help individuals and organizations navigate this transformation in practice through teaching and consultation.
Research interests — Role Design & System Prompts · AI in Education · Philosophy & Ethics of AI · Multi-Agent LLM Systems · Social Simulations · Governance & Economics of AI · Autonomy, Creativity & AI.
Education
PhD, Artificial Intelligence in Education
Tallinn University · AIED-TLU, School of Digital Technologies · Supervisor: Danial Hooshyar · 2026–present
MSc, AI for Sustainable Societies (AISS)
Erasmus Mundus Joint Master's (Full Scholarship) — Tallinn · Lusófona · Tampere · 2024–2026 · Cum Laude; thesis awarded highest grade
BA, International Relations
Eötvös Loránd University · Stipendium Hungaricum (Full Scholarship) · Thesis: Algorithmic Power & Politics of AI · 2019–2022
BA, Business Administration
KU Leuven · 2014–2018
IB Diploma Programme
2011–2013
Professional Experience
AI Consultant & Educator (Independent)
Remote · 2022–now
AI literacy, prompt & intent engineering, practical workflow design
Workshops & coaching for individuals and organizations
Recent clients: Cancer Tools UK, Perzepolis (tourism agency)
Co-Founder, IBEX
Tehran, Iran · 2018–2019
Built & sold 60+ handmade children's bicycles; owned branding, production, supply chain & marketing
Head of Global Talent Program, AIESEC Belgium
2017–2018
Led national operations for 100+ professional internships; co-facilitated 5 national conferences for 300+ members
Committee Coordinator, AIESEC in Campus Brussel
2015–2016
Led a team of 7; secured 36 internships; organised a job fair (9 firms, 400+ attendees, €20,000 raised)
Research & Projects
Designing Roles for AI Agents (Master's Thesis)
An ontology for how normative roles shape LLM-agent behavior across multi-agent simulations, applied to 38 studies. Co-authored paper published by Springer Nature (WorldCist'26).
ReaLLM
A prototype exposing the hidden instructions that shape LLM behavior — testing interface-level transparency as a safety intervention. Accepted for workshop at CPDP 2026, Brussels.
Civic AI Debaters
A web study comparing single- vs. multi-agent setups to examine how model configuration shapes bias in civic learning.
Beijing Air Quality Predictor (ML)
Python time-series analysis (EDA, feature engineering, ML modeling) plus a forecasting web app deployed via Streamlit.
AI Academy
A prototype role-playing game for kids to interact with LLM agents in different roles and learn how prompts influence behavior.
Skills
Data Science & Machine Learning
EDA, visualization & predictive ML with Python, Pandas, NumPy, Scikit-learn, Matplotlib; statistical analysis & modeling.
LLM Application Development
Designing multi-agent systems & agentic workflows with LLM APIs; prompt/context engineering and system design.
Team coordination, cross-cultural collaboration, project management.
Courses & Fellowships
Centre for AI Safety: AI Safety, Ethics & Society (AISES) · BlueDot Transformative AI Course · Goethe Institute «AI to Amplify» fellowship · Effective Thesis Accelerator · HuggingFace AI Agents Course.