Felipe Buchbinder
Felipe Buchbinder has a talent for seeing patterns where others see only chaos. Across a career spanning academia, consulting, banking, and development finance, he believes that data science is more than just tools or equations. Data science is about the perspective you bring to the world and how you use it to create meaningful impact.
Now working at BNDES, the Brazilian Development Bank, and teaching on the side, Felipe reflects on that philosophy with the ease of someone who has lived it. We sat down with him to talk about microcredit, storytelling, AI, and what he wishes he had known when he started.
When talking about the connections of his varied career experiences, Felipe explains, “Data science is actually about a way to look at the world,” he says. “It’s not about the math. It’s not about the tools. It’s more about how you approach a problem.”
That approach has taken him across wildly different domains, such as climate change, financing, microcredit, and academia, but the core skill, he argues, remains the same: understanding the problem deeply, and then bringing people along with you toward a solution.
“Very often you will be dealing with people who have no background in data whatsoever. This job of explaining and convincing, sometimes it’s almost like making them arrive at the conclusion for themselves.”
It’s a surprisingly human-centered view from someone in a highly technical field, but for Felipe, it makes perfect sense. Organizations, he notes, run on myths, stories people tell themselves about how things work. Some of those stories are true. Some are wrong but convenient. And some are just plain wrong. “That’s where data can make a big difference,” he says. “When something is blatantly wrong and inconvenient, people wouldn’t want it to be that way, they simply have the wrong ideas. That’s where you can really move the needle.”
Microcredit and the Data of Real Life
At BNDES, Felipe currently works in microcredit, a field he describes with the kind of passion that suggests he has thought about it a great deal. “About 70% of jobs come from small companies,” he explains, “both in the United States and in Brazil. We look a lot at Amazon and Google, but the truth is that most employers are local markets and local companies.”
The social dimension is what really drives him. He paints a vivid picture: a woman with several children, a husband who isn’t around, trying to build a small business selling something she knows how to make. The one thing standing between her and economic independence is access to capital.
“You give her the credit, and you change her position within the family. She is no longer the woman who stays home. She’s the person who owns a business and who brings money in.”
There’s also something surprising about gender equity in this space. “When you look at wage disparity between genders, it is smallest in the microcredit world,” Felipe says. “If I sell popcorn and you sell popcorn, there is no reason why someone would pay more for mine simply because I am a man.” Economic independence, it turns out, is one of the most effective equalizers.
When Traditional Data Doesn’t Work
Celebrating his last class for his course on Public Policies on Climate Change and Sustainability.
Working in microcredit comes with a distinctive challenge: the people you’re trying to help often don’t fit neatly into traditional data systems. They may not have formal employment records, pay taxes on their income, or have a consistent financial footprint. The woman who polishes nails one day, cuts hair the next, and sells popcorn the day after has real economic data, but it doesn’t show up in a credit bureau.
“Data that banks traditionally use for credit scoring works very poorly here,” Felipe explains. “You don’t have this data. It’s missing, and it’s messy, and even if you have it, you’re not sure you can rely on it. This is often more difficult than dealing with lots of data.”
So what does work? Behavioral data, surprisingly. “People who are very optimistic tend to have higher default rates,” he notes. “People who are skeptical, who think their business might go wrong, tend to have lower default rates.” The team also draws on social media, government datasets, location data, and even insights from neighbors and community members. “Your neighbor knows if you’re a trustworthy person,” he says simply.
The Art of the Data Story
If there’s one skill Felipe returns to again and again, it’s storytelling. “There is not a single data initiative I do that doesn’t require storytelling,” he says. “No matter what you’re doing, even a simple analysis in Excel, you will have to explain it to someone.”
He tells a story about building a model to identify credit-constrained companies. To explain it to senior executives, he didn’t lead with methodology. He picked examples: the biggest Brazilian oil and gas companies on one end, obviously flush with capital; tiny businesses without even a website on the other, obviously underserved. One company disappeared from the dataset between years because a millionaire had bought it, instantly resolving its cash flow problem.
The key, he says, is knowing your audience. Executives need an intuitive story. Technical peers need rigor. The same underlying truth requires a completely different frame depending on who’s in the room. “What works for one person will not work for another. It’s important to understand that context.”
This, he adds, is exactly what he loved about MIDS. “If I had done a master’s that was all about tools, it would have been outdated very quickly and it wouldn’t reflect the skills I actually use the most.”
Academia, AI Ethics, and the Next Generation
Outside the bank, Felipe teaches, and it’s in the classroom that some of his deepest concerns about AI surface. As he has grown older, he says, he has become less interested in the latest tools and more interested in their implications: for the job market, for inequality, for mental health, for what it means to be human.
He cites a research finding that the number one use case of AI is as a form of psychological support. “Number three is to find purpose in life,” he adds. “That really hit me. Not a significant other. Not a hobby. An AI.”
In his classes, he sees students who sit separately from people they might have once befriended, who barely talk across social lines. He sees students who learn prompt engineering primarily to use it to cheat in other subjects. And he worries about what gets lost when AI starts to handle the hard emotional work of being a person.
“I went to ChatGPT, poured everything that was in my heart, and what came out was perfect. And then next time I needed to be polite with someone, I felt the urge to go back and I thought, wait. I cannot lose the ability of being polite. I cannot lose the ability of being angry and yet expressing it professionally.”
He also raises the privacy dimension: that people in Brazil (and everywhere) routinely hand over their data (a national ID number at a pharmacy for a small discount, for example) without understanding what companies can do with it. “It might cost you a job. People don’t think about that.”
“Education has two purposes,” he says. “One is to prepare people for the job market. The other is to form good citizens such as people who can critically analyze the world. Even if you end up cultivating a garden and selling plants, you will still need to know what your data is used for.”
Advice to a Younger Self
Asked what he would tell himself at the start of his career, Felipe doesn’t hesitate long. The first lesson: don’t underestimate politics. Just because you have the truth doesn’t mean the doors will open. Pick your battles and go after the myths that are both wrong and inconvenient, not the ones that are wrong but comfortable for the powerful.
The second: talk to people before you start building. “We love data. We love mathematics. We are eager to jump into a problem and start modeling and that’s almost certain failure.” Understanding the business context, the constraints, the true problem (which is rarely the stated problem) is time that pays off many times over.
And a practical note: build your portfolio. “I wish I had done more of that as a student. Now my projects are for the bank and can’t be public, so do it while you can.”
It’s the advice of someone who has learned the best data scientists are listeners, diplomats, and storytellers.
─────────────────────────────────────────────
Felipe Buchbinder (’21) is an alumnus of the Duke Master in Interdisciplinary Data Science (MIDS) program. He currently works as a senior data scientist at BNDES, the Brazilian Development Bank, and teaches data science and AI ethics at Fundação Getulio Vargas (FGV) in Brazil.