Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
What we learned
What's next for Inspiration
Data scientists spend 80% of their time fixing unstructured, corrupt datasets. Manually writing the Python logic to fix missing IDs, broken date formats, and negative spending values takes hours. I wanted to eliminate this bottleneck using an autonomous background agent that handles the heavy lifting while the user steps away.
What it does Auto-EDA is an end-to-end autonomous data janitor. Instead of manual coding, a user passes a messy dataset to a Gemini-powered asynchronous agent. Without any human intervention, the agent acts as a senior data engineer:
It analyzes the schema errors and anomalies.
It writes the necessary Pandas cleaning script on the fly.
It dynamically executes that code in the background to repair the data.
It analyzes the newly cleaned data, determines the best way to visualize the distributions, writes the plotting code, and renders a publication-ready Matplotlib/Seaborn dashboard.
Zero manual Pandas or visualization code is required from the user.
How I built it I built this in Google Colab using Python. The core engine is powered by the Gemini 3.6 Flash API. I engineered a prompt pipeline that takes the raw dataframe's data and missing value counts as state context. It passes this context to the LLM to generate raw Python code, and uses Python's exec() function to dynamically execute the data cleaning and visualization steps autonomously.
Challenges I ran into (and why my demo is silent!) This is my very first hackathon, and I built this late into the night (around 2:00 AM!). My family was asleep, and if I made a sound to record a voiceover for my demo video, my hackathon journey would have ended right there. I had to quickly pivot to a completely silent, text-driven video presentation to explain the architecture without waking anyone up.
Additionally, as a beginner, learning how to handle dynamic code execution, API security, and Python environment bugs on the fly was a massive but incredibly rewarding challenge.
What's next for Auto-EDA I am using this project as a stepping stone to rapidly master Python engineering before I transition fully into advanced data science. My ultimate goal is to prepare for rigorous engineering programs like MIT, and building self-correcting AI pipelines like Auto-EDA is the exact foundation I want to keep expanding on.Auto-EDA: The Autonomous Data Janitor
Log in or sign up for Devpost to join the conversation.