AI_for_biomedical_students

Rapid prototyping

• 🎥 Lecture on deeplearning.ai

• ⬆️📎 Code github

Co-creating an MVP with GenAI

• 🎥 Lecture

• Use a GenAI app to prototype (turn PDF notes into clinical insight)

• Brainstorm end goal

• Who is audience? (non-technical people)

• What does the app look like?

• Features (drag and drop, export to csv, etc.)

• be specific (do not give open ended tasks to GenAI apps)

⚠️ NOTE

Give the AI a role act as a product manager

⬆️📎 Tools

🎥 Lecture

• Google Colab

• Streamlit

• ngrok account and auth token

• Sarvam AI API key

Optional

• Python

• Local files

🎮 🛠️ Activity using Streamlit

• streamlit

• Interactive streamlit GDP dashboard

• Deploy on streamlit.io

• Click on New app

• Here is a deployed app

• Push your code to github and deploy

• Course on deeplearning.ai on streamlit

• Setup a repository in github

• Or run in Google Colab

• Run streamlit using

streamlit run script.py

• Here is an example script

import streamlit as st
import pandas as pd
import numpy as np

st.write("Streamlit supports a wide range of data visualizations")

all_users = ["Alice", "Bob", "Charly"]
with st.container(border=True):
    users = st.multiselect("Users", all_users, default=all_users)
    rolling_average = st.toggle("Rolling average")

np.random.seed(19)
data = pd.DataFrame(np.random.randn(20, len(users)), columns=users)

if rolling_average:
    data = data.rolling(7).mean().dropna()

tab1, tab2 = st.tabs(["Chart", "Dataframe"])
tab1.line_chart(data, height=250)
tab2.dataframe(data, height=250, use_container_width=True)

• ngrok signup

• create an auth token

• Google Colab notebook

• To save your ngrok authtoken securely in Google Colab, you should use Colab’s Secrets feature. This allows you to store sensitive information like API keys or tokens without embedding them directly in your code, which is important for security and when sharing notebooks.

1. Open the Secrets Panel: In the left sidebar of your Colab notebook, click on the “🔑 Secrets” icon (it looks like a key).
2. Add a new secret: Click the ”+ Add new secret” button.
3. Name the secret: For your ngrok authtoken, you could name it YOUR_AUTHTOKEN (or any other descriptive name you prefer, but remember it for later).
4. Enter the secret value: Paste your ngrok authtoken into the “Value” field.
5. Save: Click “Save secret”.

Once saved, you can access this secret in your Python code using from google.colab import userdata and then userdata.get('YOUR_AUTHTOKEN') (replacing 'YOUR_AUTHTOKEN' with the name you gave your secret).

Streamlit commands

• 🎥 Lecture

• st.slider st.button st.text_input() st.selectbox() st.checkbox() st.file__uploader()

🎮 💡🛠️Activity: chatbot and streamlit

• Code to wrap a streamlit UI around chatbot

• Code from deeplearning.ai

# import packages
from dotenv import load_dotenv
import openai
import streamlit as st


# load environment variables from .env file
load_dotenv()

# Initialize OpenAI client
client = openai.OpenAI()

st.title("Hello, GenAI!")
st.write("This is your first Streamlit app.")

response = client.responses.create(
    model="gpt-4o",
    input=[
        {"role": "user", "content": "Explain generative AI in one sentence."}  # Prompt
    ],
    temperature=0.7,  # A bit of creativity
    max_output_tokens=100  # Limit response length
)

# print the response from OpenAI
st.write(response.output[0].content[0].text)

Visualization

• 🎥 Lecture video

• st.bar_chart()

• st.scatter_chart(df), works with pandas dataframe

• use with matplotlib st.pyplot(fig)

• use with plotly st.plotly_chart(fig, use_container_width = True)

Deploy app

• 🎥 Lecture video

📚 📝 Deployment for streamlit

• Github Create new account on github

• Create new app and connect to github

• Create new repository on github

• Upload your code to github (code below)

import streamlit as st
import pandas as pd
import numpy as np
import folium
from folium.plugins import HeatMap
from streamlit_folium import st_folium

st.title("Outbreak Investigator")
st.write("Adjust the settings in the sidebar, then try to identify the source of the outbreak from the map.")

# --- SIDEBAR CONTROLS ---
st.sidebar.header("Settings")
total_cases = st.sidebar.slider("Total cases", 100, 1000, 500)
cluster_pct = st.sidebar.slider("% of cases near the source", 10, 90, 70)
show_source = st.sidebar.checkbox("Reveal the true source")

# --- GENERATE SYNTHETIC DATA ---
market_lat, market_lon = 30.6195, 114.2577

cluster_count = int(total_cases * cluster_pct / 100)
noise_count = total_cases - cluster_count

np.random.seed(420)
cluster_lats = np.random.normal(market_lat, 0.005, cluster_count)
cluster_lons = np.random.normal(market_lon, 0.005, cluster_count)
noise_lats = np.random.uniform(30.50, 30.70, noise_count)
noise_lons = np.random.uniform(114.20, 114.40, noise_count)

cases = pd.DataFrame({
    'lat': np.concatenate([cluster_lats, noise_lats]),
    'lon': np.concatenate([cluster_lons, noise_lons]),
})

pois = pd.DataFrame({
    'name': ['Wuhan International Plaza', 'Huanan Seafood Market', 'Hankou Railway Station', 'Wuhan CDC'],
    'lat':  [30.584,   30.6195, 30.618, 30.612],
    'lon':  [114.271,  114.2577, 114.250, 114.265],
    'is_source': [False, True, False, False],
})

# --- SHOW STATS ---
st.write(f"**Total cases:** {total_cases} — **Clustered:** {cluster_count} — **Scattered:** {noise_count}")

# --- BUILD MAP ---
m = folium.Map(location=[30.61, 114.28], zoom_start=13, tiles='cartodbpositron')

HeatMap(cases[['lat', 'lon']].values.tolist(), radius=12, blur=15).add_to(m)

for _, poi in pois.iterrows():
    if poi['is_source'] and show_source:
        color = 'red'
        label = f"TRUE SOURCE: {poi['name']}"
    else:
        color = 'black'
        label = poi['name']

    folium.Marker(
        location=[poi['lat'], poi['lon']],
        popup=label,
        tooltip=label,
        icon=folium.Icon(color=color, icon='question-sign'),
    ).add_to(m)

st_folium(m, width=900, height=550)

• requirements.txt is here

streamlit>=1.35.0
streamlit-folium>=0.20.0
folium>=0.17.0
pandas>=2.0.0
numpy>=1.26.0

• Connect to code on github

• Go to Streamlit and connect your github repository

• Deployed app

• can also deploy on Huggingface spaces

(Optional) Data handling

• 🎥 Lecture video

• st.session_state()

This site is open source. Improve this page.