AI Agents
Build a public-health AI agent from scratch โ define its persona, train its knowledge base, choose a channel, and test it live.
Start the guide โData Projects
Hands-on with real public health datasets โ low-code analytics, visualization, and AI-driven modeling with Tableau & Orange.
Start the guide โBuilding Applications
Turn a prototype into something people can use โ vibe code a working web app and ship it to your own public URL on GitHub Pages.
Start the guide โThe Top 10 finalists are out
Ten Maker Space projects were selected to present on the final day โ competing for a full Online Track scholarship.
Submit your project abstract
Built something in the Maker Space? Submit your AI Agent, Data Project, or Web Application to present on the last day.
Enter your session code
The hands-on Maker Space guides unlock during the live session. Enter the code your instructor shared to continue.
Maker Space 1 ยท Hands-on Guide
Build an AI Agent with Jotform
In this session you’ll create a working AI agent for a public-health use case โ in small groups, from a blank canvas to a tested, shareable agent.
What you’ll do
Work in small groups. Together you’ll agree on a use case, create the agent, build its knowledge base with help from LLM interfaces, publish it, test it with real questions, and iterate.
๐ฅ Form your group & pick a topic
Get into small groups and agree on a single, focused public-health topic for your agent โ for example a rumor-clarification assistant, an emergency-health explainer, or a plain-language risk communicator.
One person creates the agent and shares their screen; everyone contributes content and test questions.
๐ญ Your agent’s persona
Before building, decide who your agent is. Agree on each of these:
Create your account & describe your agent
Go to jotform.com and create a free account (sign up with email or Google). Open the AI Agents workspace and start a new agent.
Jotform can build a first draft for you: in the “Describe your agent” box, type a short description of what your public-health agent should do โ for example, “a public health assistant that answers questions about seasonal flu in plain language.” Jotform AI creates a starting agent from your description.
Define the persona
In the Train tab, open AI Persona โ this is where you make your agent talk and act the way your group decided. Set each part of the persona:
- Name โ the name shown in the conversation (e.g. Maya).
- Role โ the agent’s job title (e.g. Public Health Information Guide).
- Tone โ how it communicates (warm, professional, reassuringโฆ).
- Length of responses โ short and plain, or more detailed.
- Voice โ the speaking voice, if you enable voice replies.
Add your content to the Knowledge Base
The Knowledge Base is what makes your agent give accurate, context-aware answers. Any information your agent should know must be added here โ paste text, upload documents, or add links.
Use LLM interfaces (ChatGPT, Claude, Gemini, Copilot, etc.) to help you draft and structure this content: summarize trustworthy public-health sources, turn guidance into plain-language Q&A, and refine it โ then add the result to your agent’s Knowledge Base.
If it isn’t in the Knowledge Base, your agent doesn’t know it. Build the knowledge base carefully โ it’s the heart of a reliable public-health agent. Always keep a human reviewing what goes in.
Explore & choose a channel
In the Publish tab, explore the channels your agent can live on โ a standalone link, a website chatbot, a voice agent, Gmail, and social channels. Pick the channel that fits your group’s use case.
WhatsApp and Instagram won’t work without a business account. For this exercise, use the standalone link or the voice agent โ they work instantly and are easy to share and test.
See results, test & share
Now test your agent. Open the share link (or make a test call for the voice agent) and ask it the kinds of questions your community would ask. Check the answers against your Knowledge Base, then refine the persona or content as needed.
When it’s working, share it with the room โ copy the public link or QR code so other groups can try it.
Maker Space 2 ยท Hands-on Guide
Data Projects with Tableau & Orange
In this session you’ll work with real, nationally representative public health data โ using low-code tools to build dashboards and AI-driven models, then interpret the results with public health judgment.
What you’ll do
Work in small groups. You’ll set up two free low-code tools, download the 2024 NHIS Sample Adult dataset, then build descriptive dashboards in Tableau and run clustering/prediction in Orange โ always checking results against the codebook.
Step 1 ยท Set up your tools
Two free, low-code tools
Tableau Public
Purpose: interactive data visualization and dashboards.
- Go to public.tableau.com
- Click Sign In โ Create an account
- Register with an email or Google account
- Download Tableau Public Desktop (free)
- Sign in through the Desktop app
Orange Data Mining
Purpose: no-code machine learning and visual analytics.
- Go to orangedatamining.com
- Click Download Orange
- Install for your OS (Windows / Mac / Linux)
- Open Orange Canvas โ no account required
Step 2 ยท Get the data โ 2024 NHIS Sample Adult
The National Health Interview Survey (NHIS) is nationally representative public health data, de-identified and approved for public use.
adult24.csv โ the datasetadult-codebook.pdf โ variable definitions๐ cdc.gov/nchs/nhis ยท Data are de-identified and approved for public use.
Step 3 ยท Getting ready for analysis
From data to insight
Describe with Tableau
Build descriptive dashboards โ distributions, trends, and group comparisons across the population.
Model with Orange
Use drag-and-connect widgets for clustering and prediction to surface patterns in the data.
Always consult the codebook
Variable names are coded โ the NHIS codebook tells you what each value actually means.
Interpret with judgment
AI results are a starting point. Read every finding through a public health lens before you trust it.
Your group’s goal
Pick one public health question the NHIS data can answer, build a dashboard that tells its story in Tableau, run one model in Orange, and be ready to explain what you found โ and what it means for population health.
Maker Space 3 ยท Hands-on Guide
Build & Ship a Web Application
In this session your group will build a functional prototype application for a public-health problem โ vibe-coded with GenAI, hosted on GitHub Pages, with your own public URL to share.
What you’ll do
Think before you build: start from the problem, not the technology. Then prototype fast with GenAI and publish to the web.
Brainstorm the problem
Don’t open a coding tool yet. First, agree as a group on what problem you’re tackling and for whom. Choose one of two methodologies to structure your thinking โ then work through its canvas together.
Design Thinking Canvas
Start from people. Human-centered โ best when the challenge is about experience, behavior, or access.
- Empathize โ who is affected? What do they feel, say, do?
- Define โ write a point-of-view statement: [user] needs [need] because [insight].
- Ideate โ diverge: many ideas, no judging. Then converge on one.
- Prototype โ sketch the simplest version that tests the idea.
- Test โ what would success look like for the user?
Data Thinking Canvas
Start from data. Data-centered โ best when you already have (or can get) data that could drive a decision.
- Question โ what public-health question should the data answer?
- Data sources โ what exists? What can you connect or generate as CSV?
- Variables & methods โ which fields matter; what analysis or AI applies?
- Decision & action โ who acts on the result, and how?
- Ethics โ privacy, bias, and representativeness checks.
Vibe code in your favorite GenAI tool
With your ideas together, start building. If your app uses existing data, have your data sources connected or CSV files generated to be consumed. If it collects new data instead โ a survey, a symptom reporter, a field-data tool โ you don’t need any data to start: the app itself will generate it. Describe what you want in plain language, look at what the AI produces, and iterate: that’s vibe coding.
Use whichever GenAI tool your group prefers (Claude, ChatGPT, Gemini, Copilot, or a dedicated app-builder).
Use this prompt template
Replace everything in [brackets] with your group’s decisions, then paste it into your GenAI tool.
You are an expert front-end developer. Build a [single-page web app] for [desktop / mobile]. GOAL Build [app name]: a tool that helps [audience] to [goal]. DATA & CONNECTIONS - Data source: [none / CSV file we generated / Google Sheets (published link) / API endpoint: URL] - Users can upload: [tabular data (CSV/XLSX) / images / text / PDFs / none] CORE FEATURES 1. [feature] 2. [feature] 3. [feature] UI STYLE - Style: [e.g. glassmorphism / material design / neo-brutalism] - Colors: [palette] - Responsive and accessible. CONSTRAINTS - One self-contained HTML file (inline CSS & JS) so it can be hosted on GitHub Pages. - Client-side code only โ no server.
Define the style of your user interface
Pick a visual direction and name it in your prompt โ the AI will follow it. Hover over any style to see a quick live example of the interface it produces:
JS libraries that elevate the experience
This step is optional โ your app works without it. If you have time, ask your GenAI tool to incorporate any of these; they add polish with one line in your prompt:
dc.js
Linked, cross-filtering dashboards โ click one chart and all the others update โ dc-js.github.io
Create your GitHub account
Go to github.com/signup and create a free account (one per group is enough โ use the account of whoever will manage the project). Verify your email before continuing.
Create a repository
Click the + in the top-right โ New repository. Give it a short, URL-friendly name (e.g. heat-risk-app), keep it Public, and create it.
Upload your HTML file
In your repo, choose Add file โ Upload files, drag in the HTML your GenAI tool produced, and rename it index.html (that’s the file GitHub Pages serves). Commit the change.
index.html.Prepare your GitHub Page
Open Settings โ Pages, set the source to Deploy from a branch โ main โ / (root), and save. GitHub builds your site in a minute or two.
main.Access the page you just created
Back in Settings โ Pages, GitHub shows “Your site is live at…” โ click Visit site to open your app in the browser.
https://your-username.github.io/your-repo/ โ share it with the room, and use it in your final-day abstract.
Your group’s goal
Leave with a working, public web application: a real problem framed with Design or Data Thinking, prototyped with GenAI, styled with intent, and live at your own GitHub Pages URL โ ready to demo and submit for the final-day presentations.