Written by
Dennis Jiang
Published on
November 2, 2024
As a founder, I love talking to customers about how they use Rosie, and a recent conversation with a bioanalytical scientist stood out. Their job focuses on analyzing whether new medicines work as intended and are safe to use. Imagine you’re part of a team developing a new drug. Before it goes to market, you need to test it thoroughly to ensure it’s safe and effective. That’s where a bioanalytical scientist comes in. They design experiments to track how the body responds to these new treatments. Recently, Rosie played a key role in helping this scientist analyze a large and complex public dataset—the Brain RNA-Seq dataset—related to brain tissue research.
The customer used Rosie to analyze a big dataset on brain cells. Think of it like working with sales data from different stores. In a sales dataset, columns represent store locations and rows represent various products. Instead of tracking sales numbers, this brain dataset tracks how active certain genes are expressed in different brain cells, like neurons and astrocytes. Both datasets measure activity across different categories—whether it’s gene activity in cells or product sales in a specific store.
In this case, the customer needed to identify “housekeeping” genes—genes that remain stable across all brain cell types. This would involve manually combing through thousands of genes to find the right ones—a time-consuming process that the customer compared to brute-forcing their way through the data.
That’s where Rosie made a difference.
The customer had tried other tools like Microsoft’s Copilot, but they didn’t work for this task. With Rosie, they asked questions in plain language, like, “What are the cell types that are stable across expression similar to this gene (HGNC:18)?” Rosie quickly analyzed the data and provided a list of relevant genes, saving a lot of time.
The impressive thing was that the customer didn’t need to explain every technical detail upfront. Rosie could infer the structure of the dataset—like which rows represented different brain cell types—without needing specific instructions. This allowed the customer to get right to the analysis instead of spending time prepping the tool.
Nope! No AI tool is perfect. The customer caught a few slip-ups, like Rosie misreading a row or spitting out the wrong value. But because they could check it against the raw data, it was easy to fix and keep going. A few hiccups didn’t stop Rosie from being super helpful.
The customer viewed Rosie as a tool to speed up, rather than replace, their work. They understood that AI can sometimes make mistakes, but it was easy to cross-check Rosie’s outputs against the dataset. They found this process more helpful and faster than if they’d done everything manually.
“I wish I had this during my PhD training—it would've been a game-changer! Most of our data came in CSV files, and Rosie would’ve made everything so much easier.”
The customer loved how easy it was to talk to Rosie. They didn’t have to use formal language or repeat basic details—Rosie just got it. This let them focus on the deeper data analysis. They said it felt more like chatting with a person than using complicated software.
Rosie helped the customer extract meaningful insights from a complex dataset, cutting through the noise and making data analysis far more efficient. The ease with which they could ask Rosie detailed, technical questions in natural language made her a valuable tool in their workflow.
While Rosie made occasional errors, the customer knew it wasn’t meant to replace their expertise. Instead, Rosie is a tool that helps speed up tasks, allowing scientists to focus on validating and refining the results with their own experiments.
For this customer, Rosie turned a slow, complicated process into something quick and easy. Rosie's ability to handle big, complex datasets and deliver fast, useful answers made a huge difference in their work.
Click here to try Rosie out yourself with our interactive demo
Thanks for reading!
Dennis Jiang
Cofounder and CEO