Data science through conversation
An AI-powered competition platform where students clean and analyse a messy dataset by giving instructions in plain English instead of writing code.
Test how you think, not how fast you type
In real data science, most of the work is understanding the data: spotting missing values, duplicates, typos and records that simply can't be true. Writing the code is the easy part.
This platform removes the coding barrier. Participants describe what they want, and an AI agent writes and runs the Python code for them. Everyone can take part, even first-semester students who are still learning pandas.
What a participant can ask
- "Give me the stats of this dataset" → runs
df.describe() - "Fill the missing values in Age with the median"
- "Remove duplicate rows and show the new row count"
- "Plot a histogram of household income"
- "What's the difference between mean and median?"
What happens when you send a message
Your message passes through four parts. The AI never runs code on our server; every command runs inside your own private sandbox.
Synthetic citizen registry records
A NADRA-style dataset with thousands of fictional citizens from Karachi, Lahore and Islamabad. All names and numbers are made up. It has been deliberately damaged: finding out how is your challenge.
| Column | What it holds |
|---|---|
Citizen_ID | Record number |
CNIC · Father_CNIC | 13-digit national ID of the citizen and their father |
Full_Name · Gender | Name and gender (M / F) |
DOB · Age | Date of birth and age in years |
CNIC_Registration_Date | When the ID card was issued |
City · Town_Area | Where the citizen lives |
Household_Income | Monthly income in PKR |
Marital_Status · Children_Count | Family details |
Hint: some problems are easy to see (missing values, typos, duplicates). Others only show up when you check whether the columns agree with each other. Think about what must be true for a real person.
Built with modern AI infrastructure
Claude AI
Anthropic's model understands each request, writes the pandas code, answers questions and judges the final session.
E2B Sandboxes
Secure, isolated cloud computers. AI-written code runs there, never on the main server.
FastAPI + LangGraph
The Python backend routes each message: code, question, score, or blocked.
Redis
Keeps track of each session, its history, the timer and the action count.
PHP Frontend
This website. It talks to the backend on the server side, so no secrets ever reach your browser.
Security first
Secret-key API, CSRF protection, rate limits, strict content policy and blocked dangerous code.
See it in action
Start a session and send your first command. Your sandbox is ready in seconds.
Access AI Agent