PROTOTYPE v0.1 · LIVE CIDMATH · EMORY

Ask an epidemiological
question. Get a
validated simulation.

EpiChat translates plain-language questions into data-informed epidemiological simulations (in Starsim) — so decision-makers can explore scenarios themselves, without waiting for a modeler.

Read how it works → Demo coming soon GitHub ↗
epichat simulation completed in 56 seconds
Simulate measles in Kenya with current vaccination coverage. layer 01 · parse → disease=measles, model=SEIR, R₀=15 layer 02 · resolve → country=KEN, MCV1=89%, contact_matrix=prem_2021_KEN layer 03 · generate → seir.py.j2 · 217 lines layer 04 · execute → n=1.2M, ensemble=False, t=56s layer 05 · narrate ↓ "At 89% MCV1 coverage, the simulation projects a contained outbreak peaking near day 94, with roughly 4.2% of susceptibles infected before herd effects dominate." ready. █
Models supported
3, more coming
SIR · SEIR · SIS
Typical round-trip
minutes
query → curve → narration
Today's wait for a modeler
days–weeks
for a policy question
Countries (roadmap)
177
contact matrices · demographics · vaccination coverage
[01 / context]
The bottleneck

The bottleneck in computational epidemiology is expertise, not compute.

During COVID-19, demand for modeling vastly outpaced supply. WHO and Africa CDC flagged this as a critical pandemic-preparedness gap — especially in LMICs with limited domestic modeling capacity.

Policy officers, surveillance teams, and health analysts need fast, defensible answers — but the modelers who can give them are a narrow bottleneck.

▸ REF · ENEN ET AL. 2026
Who can model / who needs answers
CAN MODEL
modelers
Est: ~15k
Epi modelers worldwide
CANNOT MODEL, NEEDS ANSWERS
everyone else
health analysts
surveillance officers
policy teams
local health ministries
Typical wait for a modeled answer days–weeks
EpiChat typical turnaround minutes
[02 / architecture]
Five layers

The LLM handles language. Templates regulate simulation code.

Layer 04 output · simulated epidemic curve
[03 / by example]
Natural language in. Validated simulation out.

A query, step by step.

▸ USER QUERY
"Simulate measles in a population of 120,000 with 80% vaccination coverage and R₀ of 12."
▼ EXTRACTED PARAMS
disease
measles (SEIR)
population
120,000
R₀
12 → β = 1.5 / day
vaccination
80% at t=0
horizon
365 days
network
age_structured
── extract · validate · generate · execute · narrate ──
▸ NARRATION

"With 80% coverage, the epidemic peaks near day 149, infecting roughly 1.7% of the population before the outbreak subsides."

[04 / status]
Prototype v0.1 · shipped

It already works, end-to-end.

▍ PROTOTYPE (v0.1)
Natural-language parameter extraction
Template-based Starsim code generation
Sandboxed execution + error recovery
Plain-language results narration
Editable parameter review step
SIR · SEIR · SIS models
▍ NEXT
Country-specific data (177 countries)
Ensemble simulations + credible intervals
SIRS · SEIRS · SEIAR models
50-query benchmark + validation study
Usability study (SUS, task completion)
Open release + hosted demo + preprint
○ COMING SOON · STREAMLIT · V0.1

Demo is on its way.
Check back soon.

The hosted Streamlit demo is not yet live. In the meantime, you can explore the source code on GitHub or reach out to the team directly.

View source on GitHub Contact the team
[05 / roadmap]
From prototype to research-grade tool

Seven phases, in priority order.

#
Phase
Complexity
Impact

Full data-source table and implementation notes in the docs.

[06 / team]
The team

Built at CIDMATH, Emory University.

Yuke Wang
PRINCIPAL INVESTIGATOR

Yuke Wang, PhD, MSPH

Yuke's research sits at the intersection of infectious disease epidemiology, environmental health, computational modeling, and AI. EpiChat is a prototype exploring whether LLMs can meaningfully widen access to epidemiological simulation.

Affiliation
RSPH · Emory University
Department
Global Health · Biostatistics
Annie Wang
MASTER'S STUDENT

Annie Wang

Annie is a master's student in the Epidemiology program at the Rollins School of Public Health. She earned her Bachelor's in Health Studies and Statistical Sciences from the University of Toronto. Her research interests focus on statistical analysis in public health, with experience in epidemiological research, data visualization, and applied health data science.

Affiliation
RSPH · Emory University
Department
Epidemiology
[07 / faq]
Common questions

Questions you probably have.

[08 / the ask]
Three ways to collaborate

Help us widen who gets
to do epidemiology.

▍ 01

Domain experts

Lend your eyes to benchmark queries and what a correct answer should look like for the diseases you work on.

ex · measles · HIV · influenza
COVID-19 · TB · cholera
▍ 02

Data contributors

Help integrate authoritative data streams into Layer 02 — so queries auto-load the right local inputs.

ex · contact matrices (Prem et al.)
surveillance feeds · demographics
▍ 03

Modelers for reference

Expand to other Epi modeling packages and provide validation support.

ex · Starsim · EpiModel · Mesa
outbreak · intervention validation
Reach out · yuke.wang@emory.edu →