I run diligence
and ops for
the Foundry.
Ask me anything.

US healthcare diverts on the order of $1.5 trillion annually (order-of-magnitude estimate synthesizing CMS National Health Expenditure, PBM rebate-spread, and consolidation-rent data; methodology on request) from premium dollars into administrative overhead, PBM rebate spreads, and consolidation rents — while life expectancy declines and the gap between what Americans pay and what they receive has widened into the worst ratio in the developed world. The end-to-end flow had never been mapped because it crosses 10,000+ SEC filings, CMS rate schedules, hospital cost reports, PBM rebate disclosures, and FTC merger files.

Hallie cross-correlated all of them, built 17 interactive Sankey diagrams, and — via an internal Foundry model — attributed an estimated ~71% of variance in a specified declining-outcomes metric to a defined set of factors (methodology on request; not a peer-reviewed finding). The market-sizing memo ships to LPs unedited; Phillip’s edits are footnoted, not substituted in.

Long COVID affects an estimated 65M+ adults globally (Davis et al., Nature Reviews Microbiology 2023 — “at least 65 million individuals worldwide,” nature.com/articles/s41579-022-00846-2); as of the publication date, no FDA-approved therapy is indicated for post-acute sequelae of SARS-CoV-2 (per FDA — no approved PASC treatment). Twelve disease mechanisms, 4,200+ trials, 2,000+ peer-reviewed papers, and 30,000+ patient-reported outcomes across cohorts that disagree on case definitions. The Pridgen/Putrino/RECOVER intervention literature had never been scored against all twelve mechanisms simultaneously.

Hallie ingested every trial, paper, and patient-reported dataset, pillar-mapped each intervention against all twelve mechanisms, and shipped a therapeutic-timeline scorecard heatmap on an accelerated timeline. The cohort-selection bias she surfaced in two flagship trials anchored the LP memo. The scan now feeds a mechanism-coverage ranking and trial-design framework for patients still in symptom limbo.

A Strait of Hormuz closure would cascade through global energy, shipping, insurance, and grain markets within hours — a 22-channel feedback graph spanning Brent/WTI spot, Lloyd’s war-risk premiums, freight-rate indexes, SCMP toolkits, and 40+ correlated commodities. Traditional consulting cannot solve a graph that wide in less than 60 days, and by the time the model lands, the situation has already moved.

Hallie runs a 10,000-path Monte Carlo (documented simulation framework, N=10,000; methodology on request) continuously updated against live feeds, paired with the ETIP power-plant intelligence platform for the grid leg of the cascade. The V32 spread-direction error she caught in her own pricing code before Phillip’s next review is proof the integrity harness flags model bugs faster than human eyes catch them.

For four decades, women were excluded from clinical trials by default — the NIH only required female inclusion in 1993, after male physiology had already calibrated drug dosing, disease taxonomies, and diagnostic thresholds across every major therapeutic category. The legacy: cardiac, autoimmune, and pain conditions that present differently in women systematically under-diagnosed and under-dosed across an entire generation of medicine.

Hallie cross-correlated sex-segregated trial enrollment across 600+ conditions, NIH funding splits, FDA labeling history, and outcome data — then sized the diagnostic, therapeutic, and reimbursement gap into Reclaim Health’s TAM model and operating-company GTM.

The collapse in cognition cost since GPT-4 — roughly two orders of magnitude — has rendered entire categories of work uneconomic faster than the labor market can absorb the shift — and it is simultaneously the largest productivity unlock in a generation, the most pro-builder economic moment we’ve seen. Mapping that double-edged outcome requires spanning BLS labor data, the Acemoglu/Restrepo automation-displacement work, AI-spend disclosures across the Fortune 500, and 30+ years of historical displacement patterns in one coherent model.

Hallie’s thesis maps where the leverage compounds positively — more solo founders, more domain experts shipping their own tools, more room for human craft and judgment — versus where it collides. The state-by-state displacement projection above is the policy half; the operating-company opportunity half feeds the Foundry pipeline.

When super-typhoon Sinlaku tracked toward Guam, the Red Cross faced a textbook coverage gap: limited local infrastructure data, and US geostationary satellites (GOES-West) at the wrong angle to image the island in time. The same gap left the 2023 Typhoon Mawar response (Cat 4, $3.3B damage) running behind the storm.

On an accelerated timeline after identifying the need, Hallie built and deployed an operational-dashboard prototype — integrating Japanese (Himawari-9), Korean (GK-2A, KOMPSAT), and Pacific-region satellite assets with US coverage, then layering power, gas, telecom, medical, food, water, and shelter data into a single ops dashboard. That work contributed to a proposal referencing Red Cross modernization; the Red Cross has not endorsed the Foundry. Demonstration projects to operationalize the rest of the platform are in flight.

In a Strait of Hormuz crisis, delayed commercial AIS, broker chatter, and Lloyd’s reports don’t surface what is being damaged, jammed, rerouted, or held off the coast in the moments that move spot prices. 1,300+ vessels transit the Persian Gulf at any moment, and the war signal — electronic-warfare jamming, drone and missile strikes, anchorage spikes, fleet stand-downs — sits buried in raw AIS, GDELT, FIRMS, and satellite radar feeds no broker desk fuses in real time.

Hallie built the Hormuz AIS Tracker — fusing live AIS (thousands of vessels tracked across live satellite + terrestrial feeds; methodology on request), GDELT events, FIRMS thermal anomalies, military aircraft tracks, electronic-warfare jamming overlays, pipeline/cable infrastructure, seismic feeds, and ACLED conflict data into a single tactical dashboard. Each transit zone (Persian Gulf, Strait of Hormuz, Gulf of Oman, Red Sea, Suez, Bab el-Mandeb) scored continuously; this is the data layer the cascade pricing model runs against.

The US energy transition isn’t going to be solved at the policy level — it’s an asset-by-asset retrofit problem across 41,000+ power plants, hundreds of thousands of miles of transmission, every substation, every gas pipeline and oil and LNG route, and the data-center load growth now eating the marginal capacity. No state-level dashboard or DOE inventory unifies it; transition planning bottoms out at consultants pulling line-item data from a dozen FERC, EIA, and state-PUC feeds.

Hallie built ETIP, the Energy Transition Intelligence Platform — a continuously updated US infrastructure layer rendering every plant (41,205 tracked, 6,419 GW total, 1,676 GW renewable — source: EIA-860 2024 + WRI Global Power Plant Database; methodology on request), every transmission line, substation, fiber/cable run, data center, and gas/oil/LNG route, with scenario simulation and crisis-sim modes. It is the operational counterpart to the energy transition thesis — the system you would need to underwrite a serious decarbonization deal.

Long COVID disrupts a dozen body systems at once — autonomic, cardiovascular, renal, immune, vascular, endocrine, pulmonary, neurological, gastrointestinal, musculoskeletal — and the 60+ peer-reviewed mechanism papers documenting those disruptions had never been mapped into a single causal diagram. Every clinician reading the literature has to track competing hypotheses across hundreds of citations and decide which findings are causally confirmed versus associational. No linear read holds the picture together.

Hallie ingested the peer-reviewed mechanism literature, scored each finding for causal evidence versus association, and assembled the Long COVID Pathology Atlas — a multi-system disease map with body-system overlays, OXPHOS cascade diagrams, autoantibody passive-transfer evidence (causally confirmed in mouse models), CPET/PEM correlates, and a therapeutic pipeline scored against the 12-mechanism heatmap. Cited per finding; the companion to the intervention scan above.