For foundations & mission organizations
I build models, network analyses, and evaluations for organizations whose findings have to survive scrutiny — foundations funding hard problems, research centers, and mission-driven groups working in climate, peace, governance, and international development. The method is the same standard as peer-reviewed publication; the timeline is consulting-paced. Where a result is uncertain, I say so. Where an assumption is doing the work, I name it.
What I bring
Five domains down the side, five capabilities across the top. Most engagements draw on three or more at once — the marks show where the combinations are routine, and where the work is already published and peer-reviewed.
| Domain ↓ Capability → | Systems science | AI / ML | Data science | Network analysis | Geospatial |
|---|---|---|---|---|---|
| Climate & GHG accounting | |||||
| Sustainability & energy | |||||
| Peace engineering & conflict | |||||
| Governance & anti-corruption | |||||
| International development |
Systems science is a transfer discipline — the same graph, the same feedback structure, the same provenance discipline shows up in a GHG inventory, a corruption network, and an SDG model. For a program officer, that means one practitioner can carry an evaluation across a mixed portfolio without handing it off between specialists.
Why it holds up
The academic methodology is what's rigorous; the publication cycle is what's slow. I keep the first and drop the second — so an evaluation you commission reads like something that could go to review, delivered on a program timeline.
Rigor
The methodology holds up to peer review even if nobody asks it to. A typical deliverable states the headline finding as a probability or range, then shows the assumptions, the sensitivity analysis, and where the result would change. False precision is not on offer — where uncertainty is real, it stays visible in the finding rather than getting smoothed away.
Independence
I have no stake in the original result, no product to sell you on top of the analysis, and no house position to protect. That's the point of an outside look — when a finding matters enough to fund, a replication or evaluation is worth more coming from someone with nothing riding on the answer. The mission filter is public and the refusal list is part of the offer.
Reproducible
I document methodology in enough detail that another competent practitioner could rebuild it. I default to open-source tools and hand over runnable artifacts you own outright, so the work keeps running after the engagement ends and a future grantee or reviewer can re-run it rather than take it on faith.
Documented
Provenance is the through-line: every number traceable to its source, every method you can defend when someone challenges it. Where AI assistance materially shapes a finding, the methodology section says so — the standard I hold my own work to is an answer you can trace, not one you have to trust.
Source · hover or focus16 published works listed on ResearchGate. The method for turning literature into source-traceable causal models is published. — you don't have to take the standard on my word.
Representative work
Two engagements with mission partners, each with a deeper write-up. I've kept the claims to what the work was and how it was done; outcomes in these contexts belong to the partners, not to me.
Governance & anti-corruption
A long-running project with the United States Institute of Peace and Transparency International, using graph data science in Python (NetworkX, SciKit-Network) to map shared directors, financial flows, and political appointments — surfacing the structure of a political-corruption network and the actors who hold it together. Findings were presented to the European Council on Foreign Relations.
More engagements — including climate and international-development work — are on the Work page.
Tell me the question and what the finding has to withstand — a board, a funder review, a public comment period, or another researcher rebuilding it from the data up. I'll reply within a week, and if it isn't a fit I'll usually say where else to look.