01 / Quantify
Measure the signal
Statistical models turn observations into explicit estimates, forecasts, comparisons, and uncertainty. Assumptions and validation remain visible rather than disappearing behind a single score.
Biostatistician · Decision Science · AI-Enabled Decision Systems
Build trustworthy decision-support systems.
Quantitative models measure signal and uncertainty.
Semantic AI works with evidence that does not fit neatly into a table.
Human judgment owns context, consequences, and the final decision.
I build trustworthy decision-support systems that combine quantitative models, semantic AI, uncertainty, and human judgment.
The goal is not automation for its own sake. It is better decisions from evidence—with assumptions visible, uncertainty carried forward, and human responsibility explicit.
EVIDENCE FIRSTUNCERTAINTY VISIBLEHUMAN AUTHORITY REMAINS
evidence / quantitative + semantic
uncertainty: carried forward
model authority: bounded
decision: human-owned
01 — From evidence to decision
The model is only one part of the decision.
Useful decision support has to connect what can be measured, what can be interpreted, what remains uncertain, and what a person ultimately needs to decide.
02 / Interpret
Use semantic evidence
Not every useful input is numeric. Specialized AI can structure text, reviews, guidelines, code, and other semantic evidence—but its output is treated as evidence to evaluate, not authority to obey.
03 / Decide
Keep judgment human
Evidence can change a decision without making the decision itself. Context, consequences, values, escalation, abstention, and final accountability remain explicit human responsibilities.
Operating principle
Uncertainty should travel with the answer.
The system should preserve enough evidence for a person to understand what supports a recommendation, what remains unknown, and when not to act.
Data are projections of reality.Models are projections shaped by data and assumptions.Reality is larger than both.
02 — Decision-system method
Start with the decision, not the model.
The same structure applies across forecasting, analytic review, AI-assisted analysis, and governed agent workflows.
01
Decision
What choice, escalation, forecast, review, or action does the system need to support? What would count as useful evidence?
defined02
Evidence
What quantitative and semantic evidence is available? What is missing, biased, unsafe to expose, or outside the observed data?
bounded03
Models + uncertainty
Use the right statistical, computational, or semantic models; validate them; compare alternatives; and carry uncertainty, disagreement, and abstention forward.
evaluated04
Human judgment
Present reviewable evidence and explicit limits at the point where inference becomes consequential action. The person keeps authority over the final call.
human-ownedBetter automation is useful. Better judgment is the objective.
03 — Existing work, one theme
Different systems. The same decision problem.
The projects below are not separate directions. They address different parts of the same problem: how to turn incomplete evidence into useful decisions without hiding uncertainty or handing away human responsibility.
Flagship / PAGe
Forecast what is still unfolding—and keep the uncertainty visible.
PAGe is a phase-aligned infectious-disease forecasting system. It treats forecasting as a decision-support problem: detect the emerging phase, estimate what may happen next, preserve probabilistic uncertainty, and provide evidence a person can interpret rather than a single opaque prediction.
Quantitativestatistical forecasting
Probabilisticuncertainty retained
Validatedout-of-sample evaluation
Humandecision support, not autopilot
Decision systems
evidence → actionPAGe supports infectious-disease forecasting. Which explores explicit decision and evaluation layers for calibration, abstention, model comparison, and semantic evidence. aiR supports statistical analysis while keeping assumptions and interpretation with the analyst. The SSC Review Agent applies retrieved evidence and structured rubrics while leaving final decisions to human reviewers.
Trust + control infrastructure
bounded AIDataGangeR reduces unnecessary exposure of source data through synthetic stand-ins. HMA binds evidence and approvals to workflow stages. AgentPorter constrains what agents can access and execute. clientAgent turns requests into reviewable candidate artifacts without giving agents release authority.
Reproducible evidence systems
provenance + validationOPCC and ONgeoR make public-health linkage and geography reproducible and source-documented. Cranary combines deterministic package evidence with semantic review. These systems make the evidence trail inspectable and reusable instead of leaving it implicit in one-off analysis.
AI expands what can be represented—not what should be trusted automatically.
Statistics contributes explicit assumptions, calibration, validation, and uncertainty. Semantic AI contributes a way to work with evidence that is difficult to encode as conventional variables. Workflow controls preserve provenance, boundaries, and review.
The common design question is therefore not Which model is smartest? It is What evidence supports this decision, what remains uncertain, when should the system abstain, and who has authority to act?
evidence?calibration?uncertainty?abstain?provenance?authority?
04 — Selected implementation examples
The method has to survive contact with real work.
These examples are supporting evidence of implementation capability: statistical tools, data systems, and usable interfaces built so another person can inspect, operate, and reuse them.
Demo-data boundary: These public demos are not approved channels for personal, health, confidential, or client-owned information. See the privacy notice. ONgeoR: source · documentation.
Demo-data boundary: These public demos are not approved channels for personal, health, confidential, or client-owned information. See the privacy notice. ONgeoR: source · documentation.
About
Why I build decision systems this way.
The professional theme comes from the same premise: models are representations, uncertainty is real, and consequential decisions require accountable human judgment.
Lennon Li, PhD, P.Stat.
Statistics taught me to distrust certainty.
Philosophy taught me not to mistake representation for reality.
AI made both lessons more important.
I am a biostatistician and decision scientist building AI-enabled decision-support systems. Statistical training taught me that estimates carry assumptions, observations have limits, and conclusions have uncertainty. Daoist and Buddhist thought pushed the question of representation further: a description can be useful without becoming identical to what it describes.
AI gives us powerful ways to represent, predict, synthesize, and generate. That power can also make representations easier to mistake for reality. I therefore build systems that preserve evidence, expose uncertainty and bias, and keep consequential judgment with people.
15+ yearsapplied public health
45+ outputsscholarly work
20+ traineesgraduate mentorship
Statistics
Quantify what we know—and what we do not.
Partial observations, sampling, bias, assumptions, competing explanations, and uncertainty are not annoyances around the analysis. They are part of the analysis.
Daoist influence
Description is useful. Description is not the whole.
道可道，非常道。《道德經》 / Dao De JingA philosophical influence here is the limit of naming and description: models and concepts help us reason about reality without becoming identical to reality.
Modern application, not a claim that the text was making a statistical argument.
Buddhist influence
Use representations without becoming attached to them as complete truth.
凡所有相，皆是虛妄。若見諸相非相，即見如來。《金剛般若波羅蜜經》 · 鳩摩羅什譯As a modern analogy, an observation, visualization, estimate, model output, or AI answer may be useful without being the whole reality it represents.
The sutra was not discussing statistics or AI; the connection here is an intellectual influence concerning appearance and attachment.
AI
Richer representations make awareness of limits more important.
AI can synthesize vast amounts of information and produce convincing outputs quickly. That power is useful. It also raises the cost of forgetting what was observed, what was inferred, what remains uncertain, and who is responsible.
Statistics teaches us to reason with uncertainty.Daoist thought teaches the limits of description.Buddhist thought teaches non-attachment to appearances.AI tests whether we remember any of them.
What remains unseen
The reality that can be modeled is not the whole reality.
The point is not to eliminate uncertainty. It is to see more clearly, know the limits of what we see, test alternatives, recognize bias, and remain open to what our models have left out.
Observe carefully. Model humbly. Decide consciously.