Research Areas
Four connected programs
The Institute's work spans prevention modeling, causal inference, healthcare AI, and computational neuromedicine. Programs share a common thread: whether a computational result is reliable enough to inform decisions about real patients and populations.
Computational HIV Prevention
Implementation and kinetics of HIV prevention modeled at population scale, from decision support to route-specific intervention limits.
LAI-PrEP Bridge Period: Clinical Decision Support Tool
Defines the 2–8 week "bridge period" as the critical implementation bottleneck in long-acting injectable PrEP and delivers a decision support tool validated across 21.2 million synthetic patients.
Finite Prevention Windows for HIV Post-Exposure Prophylaxis
Irreversible proviral integration defines a finite, route-specific PEP window; parenteral windows compress roughly threefold versus mucosal exposure, and structural access delays bound population-level efficacy.
Structural Barriers, Stochastic Avoidance, and Outbreak Risk in HIV Prevention for People Who Inject Drugs
A population-specific analysis of the structural, legal, and system barriers to PrEP initiation among people who inject drugs, evaluating syringe service program integration as the primary delivery pathway.
Calibration-to-Deployment Mismatch in HIV Prevention Trials
A formal derivation showing how cross-sectional incidence estimators produce systematically biased estimates under structural censoring — a program that also anchors the causation work below.
Causation and Evidence Integrity
The reliability of inference from populations to individuals, and the structural conditions — censoring, selection, and observation probability — that quietly bias it.
Calibration-to-Deployment Mismatch: How Structural Censoring Biases Counterfactual Incidence Estimates
Derives survival-biased effective MDRI and a joint IRR bias factor for the Kassanjee/Gao estimator, applied to 34 high-burden US metropolitan areas using AIDSVu surveillance data — showing interventions can appear artificially superior in the populations trials aim to serve.
Population-to-individual bounds in prevention
The Finite Prevention Windows framework doubles as a causal-reasoning contribution: it converts within-host kinetics into population-level efficacy bounds, distinguishing structural failure from pharmacological failure.
Healthcare AI and Reproducibility
Whether clinical prediction models are reported, validated, and equitable enough to deploy — and whether the venues reviewing them are equipped to judge.
Synergistic Barriers to Algorithmic Recourse and Structural Discrimination in Healthcare AI
Identifies and quantifies synergistic algorithmic discrimination — where features equitable in isolation combine multiplicatively to produce discriminatory outcomes — and proposes a recourse framework for compounding bias.
Quantitative Competence and Editorial Gatekeeping at Major US Medical Journals
A cross-sectional audit of editorial quantitative competence in medical machine-learning publishing, with modeled triage-error implications for quantitatively intensive manuscripts.
AI Readiness Framework
Six gated questions structuring how a clinical prediction model's readiness is assessed, aligned with TRIPOD+AI reporting expectations.
Computational Neuromedicine
Signal-processing and registry methods for neurometabolic states across phases of HIV infection.
Noise Correlation Length Distinguishes Neurometabolic Protection from Vulnerability Across HIV Infection Phases
A signal-processing biomarker framework, released with the first publicly available consolidated dataset of phase-specific HIV neurometabolic MRS data — 49 observations across 7 human cohorts spanning 1985–2021.
Quantum Coherence Preservation in Fibonacci-Structured Microtubules During HIV-Induced Neuroinflammation
Computational modeling of coherence preservation in microtubule geometries as a candidate mechanism for the HAND cognitive paradox. Full journal submission is held pending an extension addressing coherence at physiological temperatures.