I research how environmental, infrastructure, and governance conditions at the county level create hidden risk in critical systems, and I build the data infrastructure to make that risk visible across all 3,127 U.S. counties.
Decision-grade, fixed-scope comparisons of the counties you are weighing, for siting, investment, and policy.
See the briefs →A multi-volume study of how government delivery, environment, and civic capacity vary across counties, built on federal and state public sources.
See the research →On the gap between public data and real community conditions. English and Spanish, in person or virtual.
See speaking →You are not paying for time. You are paying for risk clarity on a decision worth far more than the brief.
This is civic and governance risk screening, meant to run before and alongside your technical engineering work, not to replace it. It surfaces the water, governance, workforce, and community-capacity risk that sits outside an interconnection study or a load model, so you walk into those studies already knowing where the non-engineering risk is.
“The county data was rigorous, sourced, and let us see siting and community-capacity risk we couldn’t find anywhere else.”
Send the candidate counties and the decision behind them, and I will scope the right tier.
amybarthelemy@govparti.orgThe independence is the point. With no stake in which county wins, the comparison can be trusted.
A brief is not a data dump. Each draws on composite indexes that turn raw county data into the questions a decision actually turns on. These also show where a community is being drained and where investment would land, not just where the risk is.
Whether a county government can function, permit, and follow through, across infrastructure, economy, health, transparency, environment, education, and civic participation.
Cumulative environmental risk in one score: environmental-justice exposure, pollution, PFAS, social vulnerability, air quality, and Superfund presence.
How reachable and usable a county government's public records are online: document-type breadth (60%) and click-depth (40%). The transparency metric, and the Transparency section of the Civic Health Score; a low score means records are hard to find or reach.
How concentrated extractive industry is in a community: payday lenders, dollar-store saturation, for-profit colleges, chain nursing homes.
The community-anchored institutions that buffer extraction: credit unions, CDFIs, legal aid, federally qualified health centers, libraries.
Upstream civic investment a community has or lacks, the things that prevent harm before it starts. Shown with a completeness caveat where source coverage is thin.
Protective plus preventive minus predatory: a single read on whether a community is being built up or drained, and where investment would matter most.
How many market-capture conditions stack in one county: hospital, broadband, electric, and banking monopolies, plus food and legal-aid deserts.
A multi-signal eviction-risk model. The full calibrated model is pending and not yet surfaced in briefs.
A county energy-burden stress score. Not yet surfaced as a calibrated figure.
One question runs through the whole program: what happens to communities and shared resources that fall through the gaps in how we govern and measure. The county is the entry point, a place whose conditions the systems above it routinely fail to see, but the same question scales down to a household missing a benefit and up to a commons no jurisdiction protects. Built solo from federal and state data sources with a graded quality-control pipeline. The series is in draft; figures below are working results, re-verified before any formal citation.
If you are a researcher or institution interested in collaborating on any of the lines of work below, or in licensing the underlying county dataset for research, I would like to hear from you.
amybarthelemy@govparti.orgSeven weighted sections plus a standalone Environment Burden Index, scored for 3,108 counties this run across a coverage set of 3,127. Section weights: Infrastructure 22%, Economic 18%, Health 18%, Transparency 14%, Environment 10%, Education 10%, Civic Participation 8%. Each component is percentile-ranked nationally, with missing components renormalized so low-coverage counties are not penalized.
Whether a federal program reaches distressed counties is not governed by application burden, navigation-heavy programs land on both sides of the outcome. What discriminates is how the benefit is rationed. Entitlements rationed by eligibility, keyed to distress, reach distressed counties; programs rationed by competitive supply, a fixed pool to the strongest applicants, fail them. A poverty-keyed refundable credit tracks poverty at +0.81, a broad income-threshold entitlement at +0.76, while a competitively awarded program runs −0.20. The proof is a within-program case: poverty-targeted by formula but claimed through an institutional application, its reach falls to essentially zero. Means-testing sets whether targeting can reach distressed places; access cost sets how much survives delivery.
EITC participation is the strongest external correlate of the Civic Health Score, at r = -0.65, a strong relationship across more than three thousand counties. Poverty rate follows at r = -0.61. The EITC is the best single annual county-level poverty signal in the dataset.
The platform documents cases where two or more federal sources publish disagreeing values for the same county-level measure, between survey-based and model-based estimates. These are not errors but documented methodology differences. The pipeline surfaces the disagreement rather than silently choosing one source.
Counties with military installations show higher news-desert scores. The pattern points to a documentation gap: incidents at installations, suicides, training accidents, toxic exposures, domestic violence, occur in counties whose local press infrastructure has thinned, so they go underreported. The effect is visible only at the county level, where the installation and the missing newsroom sit in the same place.
This is early, county-level work, not a settled estimate. The underlying disenfranchisement measure is preliminary: parts of it rest on proxy and incomplete inputs, so the magnitude is not yet publishable as a hard figure. It is presented as a direction worth investigating, not a verified result.
A wider agenda in development under the GovParti Research Collective, spanning dozens of papers. Each theme is one face of the same question: who and what gets left out of the systems meant to see them.
The F206 framework across federal program types: how the rationing rule, eligibility versus competitive supply, decides which programs reach distressed counties and which fail them, plus independent reconstruction of county employment estimates.
Composite-index methodology, what standard civic-capacity measures capture and miss, the disagreements between federal sources, and the intelligence engine itself as research infrastructure.
An archival inventory of federal datasets that have disappeared, changed, or degraded, and what their loss means for anyone trying to measure communities over time.
How large infrastructure loads interact with county rent pressure and energy burden, the private vendors that mediate public-records access, and the environmental exposure pathways that follow.
The placement geography of hospitals, prisons, federal facilities, higher education, major employers, and religious institutions, and what those siting patterns reveal about a community.
Compound civic-infrastructure failure in the counties where military families live, and the news-desert mechanism that keeps it undocumented.
Academic-publisher contract disclosure, private-equity consolidation in K-12 ed-tech, and the price, citation, accessibility, and retraction patterns that shape what the public can actually read.
The Caselore trajectory: a citation-annotation instrument that doubles as a teaching tool, and what curricular features predict legal reasoning.
Indigenous data sovereignty, CARE-principles posture, and platform design that supports community-controlled civic life.
The county is one example of a community that falls through governance gaps. The frame extends: the orbital and lunar commons, and the atmospheric and marine exposure pathways of large-scale buildout, are shared resources that no jurisdiction adequately measures or protects. These are governance and policy arguments rather than county-data papers, the same question of the ungoverned, carried past the county line.
A separate thread of political-philosophy work on recognition and authority runs alongside the empirical program.
Public-interest support for reporters and newsrooms: how to get the documents, and the county-level data that gives a local story its national context. You stay the requester; I bring the expertise.
Request drafting, agency targeting, scoping to avoid fees, and fee-negotiation strategy. You file in your own name, which protects your news-media fee status; I supply the targeting and the language.
County-level civic, environmental, and infrastructure data to ground a story, with sourcing a fact-checker can follow. The number that turns a local anecdote into a documented pattern.
The data surfaces the anomaly; the records strategy gets you the paper trail that proves it. Worked together, keyed to your story and your deadline.
Public-records law gives journalists reduced or waived fees, but a paid third party filing on your behalf can be classed as a commercial or middleman requester and lose that status. So the records you need are filed by you, in your name. I provide the drafting, the agency targeting, the scoping, and the negotiation playbook behind it.
The county, the agency, or the pattern you suspect.
amybarthelemy@govparti.orgTalks across technical, public-health, and infrastructure-security audiences. English and Spanish, in person (travel and lodging covered by the host) or virtual.
On civic data, infrastructure siting risk, military communities, and government transparency. Broadcast-quality remote audio from a home studio, no setup friction. Honoraria welcome where budget allows.
Research narration: I produce audio versions of papers and reports for accessibility and reach. Professional narration from the same studio, for researchers who want their work heard, not just read.
Most engineering effort goes into getting data in, and once the rows land, we trust them. This talk is about everything that happens after that, and why the post-write layer is where correctness actually gets decided. The case study is GovParti, built solo with no team to catch mistakes downstream, so governance had to be engineered into the schema and storage rather than enforced by process. The closing lesson: a validation layer that fails silently is worse than none, because it manufactures confidence you did not earn.
When critical infrastructure is proposed for a county with water contamination, no community impact assessment requirement, and a large infrastructure gap, the question is not whether the facility is secure. It is whether the community surrounding it can sustain it. This session presents findings from a civic intelligence platform analyzing all 3,127 U.S. counties, and a framework for evaluating civic readiness as a factor in infrastructure siting risk: the water, power, governance, and community-capacity layers that determine whether a facility can operate reliably over its full lifecycle.
Military communities face a paradox of civic invisibility: service members and families reside in counties where installation jurisdiction excludes them from democratic participation while they experience systematic gaps in health surveillance and social service access. This talk presents a county-level military civic disenfranchisement index, the first dataset to quantify this pattern at scale, and what it means for targeting public health infrastructure to the installation communities most in need.
What can one developer actually build with modern AI tools, an open-source stack, and a clear problem to solve? A practical walkthrough of what it takes to build AI-powered infrastructure at scale without a team, without enterprise tooling, and without a large budget: the architecture decisions, the AI integration that powers county-level civic briefings, and the data-resilience design that keeps the platform running even when federal data sources go offline. The central lesson is about what these tools make possible right now for solo builders and small teams.
Two linking problems that look identical but are governed by opposite rules: exact-key citation resolution versus probabilistic entity matching, and why a false merge is catastrophic while a missed match is merely incomplete. On auditing the canonical source you match against, and finding its gaps sit exactly where the most interesting actors are.
A county budget line in isolation is trivia; cross-referenced against ten years of history, peer counties, and dozens of federal datasets it becomes a finding. The four-layer civic intelligence pipeline behind that, and the validation discipline required when being wrong has civic consequences.
Software grows like a snowball, piling up until nobody can read its shape, or like a spiderweb, where every strand is a connection. Reading the relationship web of a real 2,400-file system to find the few threads that signal rot, which matters double in agent-written code that lands faster than humans review it.
An honest account of solo-operating a production platform as the only human who reviews what the agents produce. What agentic workflows genuinely handle, the failure modes to expect, and why verification had to become its own engineering practice. The role did not get easier; it moved.
I am an independent researcher and the founder of GovParti, a civic intelligence platform covering all 3,127 U.S. counties across environmental, health, infrastructure, workforce, and governance indicators. GovParti is one part of a broader system I build and maintain, alongside Caselore, Iris, Aegis, and the Living Almanac, so the data, the tooling, and the public archive reinforce one another rather than standing alone.
My work sits at the seam where public data meets real community conditions, the place where eligible people miss benefits, where infrastructure gets built on fragile ground, and where the counties least able to absorb risk are the ones least visible in the data. I build the infrastructure that makes those gaps measurable, and I write about what it reveals.
Or the room you need a speaker for.
amybarthelemy@govparti.org