Selected work
One line I keep coming back to: the world's most-polluting 5% of power plants produced about 73% of electricity-sector CO2 in 2018, a finding I co-authored. Engagements span academic research, government, NGOs, and enterprise. Each project sits in more than one domain, and draws on more than one kind of work. Below: three representative engagements.
Case studies
Work with the sources attached
Network analysis of political corruption in Bosnia
United States Institute of Peace · Transparency International · 2021–2025
A graph-data-science project mapping the structural relationships between political parties, state-owned enterprises, and corporate actors in Bosnia and Herzegovina. I built a custom data pipeline in Python (NetworkX, SciKit-Network) to clean, link, and analyze records on shared directors, financial flows, and political appointments. Applied centrality measures — betweenness, PageRank, degree — to surface the hubs of influence within the political-economic network. The findings went to the European Council on Foreign Relations as a presentation and a short report.
Targeting hyper-polluting power plants
RASEI, CU Boulder (NSF-funded) · 2019–2021
The world's most-polluting 5% of power plants produced 73% of electricity-sector CO2 in 2018. I was the data engineer on the University of Colorado team behind that finding: I geocoded the plants that had no location, reconciled the records across three plant databases, and matched each plant to the policies that apply to it. Published in Environmental Research Letters (Grant, Zelinka and Mitova, 2021).
Systems thinking for Burkina Faso judicial reform
United States Institute of Peace · 2019–2021
Multi-year engagement supporting the design and implementation of systems-thinking training for Burkinabé judicial reform. I worked on the method with USIP's team rather than importing a fixed curriculum, and USIP trained the Burkinabé facilitators at CIFDHA who ran the six working sessions — with judicial staff, security services, local government officials, and community representatives. I designed the method and the tools, was in the room for one session, and synthesized the resulting qualitative data using grounded theory and causal-loop diagramming to identify performance and trust bottlenecks. USIP published its own account of the Manga process and credits my contribution to it by name.