.jpg)
Blue Ridge Data Science Institute ("Bird's Eye")
Computational science rooted in Appalachia — applied to health systems, environmental conservation, and public policy across the nation.
To calmly transform chaotic real-world data and systems into flowing, nimble insights.
What We Do
BiRDSI applies rigorous computational science across three intersecting domains — each grounded in real-world data, each serving communities that matter.
Predictive modeling, epidemiology, biostatistics, and cost-effectiveness analysis applied to complex clinical questions. Specialty expertise in immunology, rare disease, and real-world evidence generation.
Data-driven analysis in defense of Appalachia's natural heritage — forests, waterways, and public lands. Quantitative evidence for conservation policy, land use decisions, and environmental advocacy.
Turning public data into public arguments. We build the quantitative case for policy positions — from federal land management rules to health equity legislation — giving advocates the evidence they need.
Featured Case Study
The 2001 Roadless Area Conservation Rule protected nearly 58.5 million acres of National Forest lands from road construction and resource extraction. Its proposed rescindment threatened one of the most significant conservation achievements in modern US history — including millions of acres of Appalachian forest critical to water quality, biodiversity, and community wellbeing.
BiRDSI conducted a quantitative analysis of the ecological, hydrological, and community impact of roadless area rescindment across Appalachian National Forest units — translating complex geospatial and environmental data into plain-language evidence for public comment and advocacy use.
The analysis drew on USDA Forest Service data, watershed delineation models, and community health metrics to build the quantitative case for roadless protection — the kind of rigorous, data-driven advocacy that turns a public comment into a compelling argument.
Seven national forest groupings in the core southern and central Appalachians. Area symbols are proportional to inventoried roadless acreage. The George Washington & Jefferson (412k ac) and Monongahela (180k ac) contain the largest roadless areas; the Daniel Boone (3k ac) the least. Schematic only — not a legal boundary map.
Roadless land is steeper than harvested land in every forest, but harvest has occurred on steep ground in several. Since FY2001, 47% of harvested acres in Pisgah-Nantahala and 42% in Cherokee were above 35% slope — making steepness alone a weak predictor of what the agency has treated in practice. Allegheny shown bold; other forests stacked behind.
Of Appalachian roadless acres classified as "likely operable" under the DEIS screen — 55% as "operable but complex" and only 14% as not operable.
People served by surface-water intakes in watersheds containing Appalachian roadless land — directly at risk from increased sedimentation.
Higher landslide density in roadless areas vs. other national forest land in the Hurricane Helene impact zone (0.48 vs. 0.10 per 1,000 acres).
61% of roadless acres exceed 35% slope, yet the DEIS operability screen classifies 86% as operable or complex — terrain alone does not protect most of this land.
Since FY2001, 47% of harvested acres in Pisgah-Nantahala and 42% in Cherokee were above 35% slope — making steepness a weak predictor of agency practice.
454,000 people are served by surface-water intakes in watersheds containing Appalachian roadless land. Tracing downstream reaches ~1.8 million served.
Hurricane Helene landslide density was ~5× higher in roadless areas (0.48/1,000 ac) vs. other national forest land (0.10) — these are failure-prone slopes.
Shifting the DEIS operability thresholds by just 10 slope points moves the "likely operable" share from 20% to 42% — the method is highly sensitive to undisclosed inputs.
All code, derived tables, and figure scripts are openly archived. Running make_all.sh regenerates every number and figure from public source data.
Analysis covers 411 inventoried roadless area polygons across 7 Appalachian national forest groupings. All code is deterministic and openly licensed. Submitted to USDA Forest Service Docket FS-2025-0001, September 27, 2026.
Who We Are
Dr. Rider is a physician-scientist, immunologist, and data scientist based in Forest, Virginia — in the shadow of the Blue Ridge Mountains he has spent his career studying and defending. He is a Professor at the Virginia Tech Carilion School of Medicine and founder of the CHILI Lab (Computational Human Immunology Lab and Innovation Hub).
BiRDSI emerged from a simple conviction: that the same rigorous computational tools used to understand rare disease and health systems can be turned toward the land, the water, and the communities of Appalachia. The mountains don't need poetry — they need data.
BiRDSI's commercial consulting arm — providing health data science, AI/ML, epidemiology, and cost-effectiveness services to pharma, health systems, and public agencies. Zanshin (残心) — the calm, sustained awareness that follows action.
Core Capabilities
Six technical domains, applied across health, environment, and policy.
Clinical NLP, EHR phenotyping, imaging informatics, and decision-support systems.
Gradient boosting, neural networks, and interpretable ML on complex real-world datasets.
Disease burden, disparities research, rare disease surveillance, and cohort development.
Survival analysis, Bayesian methods, causal inference, and regulatory-grade statistical analysis.
Markov models, budget impact analysis, ICER, and payer-facing value evidence.
Land use modeling, watershed analysis, and environmental impact quantification for policy advocacy.
Representative Work
Across health, environment, and policy — a sample of BiRDSI's analytical work.
Machine-learning pipeline identifying undiagnosed PI patients from structured and unstructured EHR data across a large integrated delivery network.
Population-level epidemiological analysis characterizing diagnosis delay, comorbidity burden, and geographic disparities in the US.
Quantitative impact analysis of proposed National Forest roadless area rule rescindment across Appalachian forest units — built for public comment and conservation advocacy.
Decision-analytic Markov model comparing IgG administration routes from a US payer perspective; submitted to support formulary review.
End-to-end statistical design and analysis support for a Phase II IIT in rare immune-mediated disease, including SAP and interim analysis plan.
Work With Us
Whether you need a rigorous analysis, an expert voice, or a partner to build something new — we'd love to hear from you.
BiRDSI · Forest, Virginia
birdsi.org · zanshininference.com