I Built a Data Analysis Tool in My Dorm. Two Professors Use It.

It started in an economics lecture.

My professor was trying to show us a scatter plot of GDP vs. life expectancy for a class exercise. He had the data in a CSV. He opened Excel. He selected the wrong columns. He selected the right columns but in the wrong order. He made a bar chart instead of a scatter plot. He deleted it. He made a scatter plot but the axes were backwards. He spent about 20 minutes on what should have been a 30-second task while 80 students watched in polite silence.

I remember thinking: this is insane. This guy has a PhD. He's brilliant at economics. He just can't make Excel do what he wants because Excel is not designed for people who think in questions instead of formulas.

The idea

What if there was a tool where you could just upload a dataset and say "show me GDP vs life expectancy as a scatter plot" and it would just... do it?

I know, I know. ChatGPT can do that. Kind of. You can paste data into ChatGPT and ask it to analyze stuff. But ChatGPT can't actually run code and show you the output. It writes code that looks right, and then you have to go run it somewhere else and hope it works. Half the time there's a bug because ChatGPT hallucinated a column name that doesn't exist in your data.

I wanted something that could see the actual data, write code against the real column names, run it, show the result, and let you keep going from there. All in the browser so you don't need to install Python or R or anything.

The build

I won't bore you with the technical details (ok fine, briefly: Next.js frontend, WebR for R execution in the browser, DuckDB-WASM for fast data queries, and an LLM agent that writes the R code). The point is it took me about three months of building nights and weekends to get something that actually worked.

The first version was terrible. The AI would write code that crashed half the time. The plots looked ugly. The data table was slow. But it was usable enough that I could show it to the professor whose lecture started this whole thing.

The feedback

He tried it with his grade data from last semester. Uploaded the CSV, asked "what's the average exam score by section", got a table in about 5 seconds. Then asked "show me the distribution as a box plot" and got a ggplot2 chart.

His reaction was something like: "oh. That's... yeah, that would have saved me a lot of time."

He's been using it since. Another econ professor in his department tried it too. That's my entire user base right now. Two economics professors and me.

What I've learned so far

Building the tool was the easy part. Getting people to try it is the hard part. Turns out "hey I built a free tool you might like" is basically the same as "hey I'm about to waste your time" in most people's mental models. Teachers are busy. They're not looking for new tools. They've been burned by edtech that promised to change their life and then required 6 hours of setup and a district-level procurement process.

The thing that's actually worked is specificity. "I built a tool where you can upload your grade CSV and ask questions about it in plain English" gets way more interest than "I built an AI-powered data analysis platform." Nobody cares about platforms. People care about solving the specific problem they have right now.

The other thing I've learned is that the "data stays in your browser" pitch matters way more than I expected. Multiple professors have told me they can't use ChatGPT for student data because of FERPA. Having everything run locally isn't just a technical choice. It's what makes the tool usable for education at all.

What's next

I need more users. Not because I'm trying to make money (it's free), but because I need feedback from people who aren't economists. Does this work for psychology researchers? Biology teachers? Political science? The data analysis part should generalize, but I have no idea what the actual pain points are for people outside the two departments I have access to.

If you teach anything that involves data and you're tired of Excel, I'd genuinely love for you to try it and tell me what's broken. I'm a college student, I'm building this mostly by myself, and I learn more from one honest "this doesn't work because..." than from a hundred polite "cool project!" responses.