A Day in the Life at Upstart: Machine Learning is Product

A Day in the Life at Upstart: Machine Learning is Product

At Upstart, our mission to improve access to credit is powered by a core belief: that artificial intelligence and machine learning (ML) can transform the lending industry. But what does that look like in everyday work? We caught up with four of our Machine Learning team members - Arman Guerra, Joe Orsini, Jon Michel, and Justin Wong to give us a peek into a day in the life of Machine Learning at Upstart.

Arman Guerra: Senior Data Scientist

💬 “We’re doing cutting-edge work that’s actually helping real people. That’s rare.”

🚴‍♂️ Daily Ritual: Biking between coffee shops, libraries, and home to stay fresh.

Joe Orsini: Director, ML Applications

💬 “ML is the product here. Our work drives strategy and business results in a very real way.”

🎵 Theme Song: “Caught in the Treads” by Gatecreeper

Jon Michel: Senior Manager, Machine Learning

💬 “We’re always experimenting. It’s fun when new techniques actually pan out—and even more fun when they improve our models.”

🐶 Fun Fact: Builds his schedule around his ‘fairly lazy’ dog’s nap routine.

Justin Wong: Manager, Machine Learning

💬 “Upstart lets us solve challenging problems that have a positive impact. It’s hard to ask for a better combo."

🧠 Known for: Writing notes to his “future self” at the end of each day to jumpstart the next one.

Why Upstart?

Everyone on the team has a different path to Upstart, but they share a common motivation: doing meaningful work with people who care.

“I’m usually skeptical when companies call themselves mission-driven,” says Arman. “But I asked every interviewer, ‘Do you feel like you’re making people’s lives better?’ And they all had thoughtful, honest answers. That sold me”​.

Justin was inspired by the blend of creativity and impact. “Upstart gives you the freedom to design solutions—and the problems we solve actually improve peoples’ lives. That combo is hard to beat”.

And for Jon, the energy comes from the work itself. “Improving access to credit is a big goal, but day-to-day, what motivates me is how fun the work is. We’re always testing, iterating, and learning”​.

Building What Matters

Upstart’s ML team is focused on a complex problem: making credit more accessible, accurate, and fair.

“Our models are what make Upstart valuable,” says Arman. “They’re more effective, inclusive, and efficient than traditional lending mechanisms”​.

Every model improvement can mean a lower rate for a borrower or open up access to someone who was previously declined. “There’s no shortage of ideas to chase,” says Justin. “The hard part is picking the ones with the most leverage”​.

And the challenge doesn’t stop at innovation—it’s also deeply technical. “It can take five years to observe the full impact of our ML decisions,” says Joe. “That makes model evaluation a fascinating statistical problem”.

Some of the team’s biggest breakthroughs include:

• Payment Transition Modeling: Predicts the in-between stages of borrower delinquency using modern ML and classical stats. “It’s a more realistic and sophisticated way to model borrower behavior,” says Jon.
• Default Risk Verification (DRVM): Surfaces hidden borrower risk and adjusts pricing accordingly. “It’s already saved borrowers money and driven $3M in revenue,” says Joe​.
• Fraud Detection: Arman helped build a new model from scratch to combat identity misuse in auto loans, “We believed we could outperform a vendor—and we did”​.

And sometimes, big ideas start in Slack. “We have a Slack channel titled ‘Machine Learning Bad Ideas of the Day’ and it is one of the best parts of our culture,” says Justin. “Anyone can pitch ideas, and some turn into real improvements”​. It’s a space where learning from failure is encouraged, and where even not-quite-right ideas can move us forward.

Life at Upstart

The digital-first model gives Upstarters the freedom to structure their day around how they work best. Arman starts his mornings at a coffee shop, works through midday at a library, and wraps up from home. “That movement keeps me fresh and focused,” he says. For Jon, it’s all about balance: “I log in early, check dashboards, then walk my dog mid-morning. Being home makes the work feel sustainable”. And for Justin, “Digital-first let me relocate with my partner for her residency—and it means we can hire the best people, regardless of geography”.

Learning and growth are core to the culture. “Upstart gives you room to define your role and expand it,” says Joe Orsini. “I’ve had the leeway to shape my team and help others grow their scope and impact”​. Jon adds, “We rotate through projects in ML, stats, engineering, and product. It’s fast-paced and multidimensional—perfect for learning quickly”.

Collaboration is at the heart of every successful team and the Machine Learning team is no exception. “I work closely with ML, product, and growth teams,” says Justin. “Each brings a unique perspective, and together we build stronger models”. That alignment is intentional, Joe notes: “There’s no cross-team rivalry—just a shared focus on impact”​.

The team’s work is grounded in Upstart’s values -

• “Every second counts,” says Justin. “There are endless ways we could improve our models—so we focus hard on what matters most”​.
• “Speed is a habit here,” Jon adds. “You’re not just analyzing—you’re building and shipping. That momentum is empowering”.
• Humility rounds out the picture. “Be Smart and Know You Might Be Wrong—that one really resonates,” says Joe. “It’s at the heart of good ML research: confidence in ideas, and openness to changing course”​.

Build the future of credit with us!

“We’re actively hiring across Machine Learning Engineering, Data Science, and Research Scientist roles,” says Justin. “Our mission is ambitious: better differentiate risk to expand access to credit. We’re looking for curious, driven people who want to make a positive mark in the world”​.

What helps someone thrive at Upstart? “Be unapologetically curious,” Justin adds. “The best ideas come from asking the right questions”.

Joe encourages future applicants to lean into rigor and creativity: “We really value foundational understanding and novel ideas. If you’re someone who not only applies models but understands why they work—and how to make them better—you’ll do well here”​.

And growth is a given. “Upstart isn’t one-size-fits-all,” says Arman. “Whether you come from a traditional stats background or something totally unexpected, what matters most is how you think—and how willing you are to learn and explore”​.

If you’re excited about applying AI to real-world problems, growing alongside a thoughtful and driven team, and helping make credit more fair and accessible, check out our open roles!