From building AI models in Nepal’s fintech scene to leading GenAI projects at a U.S. healthcare giant, Shalom Bhatta has seen the full arc of machine learning evolve—and he’s bringing that wisdom straight to this year’s hackathon. In this spotlight, he breaks down his journey, lessons, and why he’s excited to help Nepali students take bold steps into AI.
Shalom’s AI journey started over six years ago in Nepal, where he worked as the sole data scientist at Khalti, a digital wallet company.
“It was just me figuring things out—building models for predictive analytics based on user behavior,” he says
That solo hustle led him to the U.S., where he earned his master’s in AI from Northeastern University, diving deeper into deep learning and computer vision.
From there, he worked at an MIT-based startup for over two years, building behavioral recommendation systems—tools that nudged people to make smarter transportation choices.
Now? He’s at Aetna, working on deep learning and GenAI models to improve customer experience scores in healthcare. In his own words:
“We use AI to analyze customer surveys. Based on that, the government allocates funding—so the stakes are high.”
Gif by abcnetwork on Giphy
📘 2. What Lessons Have You Learned Along the Way?
Shalom didn’t hold back on this one.
First, a bold take:
“The way we’re chasing artificial general intelligence through generative AI? It’s not the right path.”
He believes math is the real foundation of machine learning—not hype.
“Linear algebra, probability, statistics—these have been around for 100+ years. Machine learning just makes them useful.”
So if you’re serious about AI? Start with math. That’s where the real edge is.
Giphy
🤝 3. What Made You Say Yes to Mentoring This Hackathon?
Turns out it was a mix of friendship and curiosity.
“Honestly, one of my best friends asked me,” he laughs. (Shivam)
But deeper than that, Shalom is excited to see how good Nepali students are at AI, to connect with the next generation, and to help demystify what it means to work in this field professionally.
Gif by southpark on Giphy
“It’s not just about models—it’s about cleaning messy data, building pipelines, deploying to web, and explaining it to stakeholders. I want students to see the full picture.”
🔍 4. What Do You Hope to See From Participants?
More than fancy models, Shalom wants to see real-world thinking.
In his words:
“Predicting is the easy part. The hard part? Cleaning data. Feature engineering. Thinking through deployment.”
“If you just call a package like XGBoost, great—but how did you prepare the data? That’s where it gets interesting.”
“I want to see how teams deal with messiness—missing values, bad inputs, large datasets—and still make it work.”
He’s also looking for model variety and thoughtful deployment plans:
“Don’t just stop at prediction—think: where does this model live? How expensive is it to run?”
Gif by AllBetter on Giphy
💬 5. Any Final Advice to Participants?
Shalom’s golden rule: start simple.
“If linear regression solves the problem, use that. You don’t need a neural network for everything.”
In other words, avoid complexity for the sake of it. Solve the problem first—then layer in sophistication if you need it.
He says it best:
“Simplest model for the simplest use case. Always.”
With mentors like Shalom in your corner, this year’s Nepal-US AI Hackathon is more than just a coding sprint—it’s a real peek into how the world’s best data scientists think, build, and solve.