EDA on a Public Dataset
Part of the AI Engineer Roadmap
In short
This project proves you can look at unfamiliar data and say something true and useful about it. Recruiters read EDA notebooks to see how you think, not just whether you can code.
A complete exploratory analysis of a real public dataset — your first project that demonstrates data intuition, not just code.
Think of it like this
A detective's case board
A good detective doesn't just collect clues. They pin them on a board, connect them with string, and end with a theory. Your notebook is the case board: each chart is a clue, each finding is a string, and the summary at the end is your theory.
Watch it happen
The project plan
1/5Choose something you care about: IPL matches, air quality, used car prices. Interest shows in the write-up.
The big ideas
What you need to know
Questions first
Three sharp questions give the notebook a story. Without them, it's a pile of charts.
Markdown between code
Every chart gets a sentence under it saying what it shows. Readers skim the text, not the code.
State the limits
Say what the data can't tell you, like 'correlation, not cause' or 'only 2019 to 2023'.
Make it easy to open
Clear outputs saved in the notebook, a README with the top three findings, and a link to the dataset.
See it in code
Walk through the code
1import pandas as pd2 3def profile(df: pd.DataFrame) -> pd.DataFrame:4 """One row per column: type, missing share, unique count, example."""5 return pd.DataFrame({6 "dtype": df.dtypes.astype(str),7 "missing_%": (df.isna().mean() * 100).round(1),8 "unique": df.nunique(),9 "example": df.apply(lambda c: c.dropna().iloc[0] if c.notna().any() else None),10 }).sort_values("missing_%", ascending=False)11 12df = pd.read_csv("your_dataset.csv")13print(df.shape)14profile(df)1/4A small helper you can reuse in every project.
A reusable profile table is the best first cell in any EDA notebook. Swap in your dataset's file name and you'll see every column's health at a glance.
Save yourself the pain
Common mistakes
Oops
Picking the Titanic or Iris dataset that every tutorial uses.
Do this instead
Pick a less common dataset from Kaggle, data.gov.in or a sports site. It stands out more.
Oops
Ending the notebook with a chart and no conclusion.
Do this instead
Close with a short findings section: what you learned, how sure you are, and what's next.
Your mission
First milestone: questions and profile
Quick check
Your EDA shows cities with more hospitals have more deaths. What's the right conclusion to write?
Go deeper
- 1Pick a dataset from Kaggle or UCI and write a full EDA notebook.
- 2Include at least five non-trivial insights, each backed by a chart.
- 3Write a short summary of what you'd model next and why.
More projects at this stop
Interactive Data Dashboard
Stuck on this topic? Ask an Insider
Get 1:1 guidance from people who've walked this exact path — on the InsideEdge app.
