Featuring: Suraj Sahu, Ph.D. student in Physics at UC Merced
Reimagining Biology Through Physics
When Suraj talks about his research, he makes you see physics not as the study of lifeless particles, but as a living, breathing language that explains how cells organize themselves, how blood vessels form, and ultimately, how life takes shape. “I study the physics aspects of biological systems,” he explains. “Right now, I’m modeling how vascular systems develop in embryos, how those first blood vessels know where to grow.”
His tools are entirely digital. No pipettes, no lab coats, just code. “Most of my day goes into writing code to understand biological systems,” he laughs. “I don’t go to experimental labs.” But coding wasn’t always easy. “I was bad at it,” he admits. “I learned during my master’s, and it was really hard, especially with my learning disability. The syntax, the symbols - it was all so abstract.”
Then came 2023, and with it, ChatGPT. “That changed everything. I could just describe what I wanted. Say, ‘Give me a code for Brownian dynamics’, and it would write one. Suddenly, I could think algorithmically instead of worrying about punctuation errors.”
AI as a Research Partner
AI didn’t just make his work faster. It changed how he thought. “When you describe a problem in words, you see where your thinking is vague. It’s like a mirror. It exposes holes in your logic. That’s when I realized AI can make you a better scientist, not just a faster one.” Still, he’s wary of overreliance. “If you let AI do everything, you lose the muscle memory of coding. Sometimes it gives me code that works beautifully, but I have no idea how. That’s dangerous.”
These days, Suraj uses AI differently. He builds his code in stages, refining each part instead of letting the model take over. He’s even moved beyond the browser window, using integrated “AI agents” inside his coding environment. “Now it can read my data folders, create code files, track every change. If I forget what I did six months ago, it reminds me like a lab notebook that never forgets.”
That notebook analogy fits perfectly. Suraj’s use of AI is meticulous. For him, it’s about efficiency, not shortcuts. “I went from spending 20 hours running a simulation to 10 minutes. That’s not replacing my work, that’s freeing me to think.”
When it comes to writing, though, he’s careful. “I only use AI for grammar. I don’t let it write the science,” he says. “I’m still learning how to write papers, and I need that struggle.”
Responsible Innovation
On the teaching side, as a TA, where AI has already become part of classroom management, Suraj has faced the usual challenges, but addressed them masterfully. “Some students just paste AI answers, even leaving in the system’s tags,” he laughs. “So instead of banning it, I asked them to use AI in front of me.”
That small change flipped the dynamic. “It made them realize how often AI is wrong, and that trust has to be earned, not assumed.”
His advice to the next generation of scientists, those who’ll grow up with AI as a given, is simple: “Don’t skip the basics. Learn what’s happening under the hood. Otherwise, you lose control of your science.”
And if he could change one thing about how universities handle AI? “Encourage experimentation,” he says. “Let students play with it safely, in structured ways. Teach them how to write good prompts. And always, always verify.”
As we wrap up, he smiles. “My relationship with AI has evolved,” he says. “At first, I trusted it blindly. Now I test it, challenge it, I argue with. That’s how science should be.”
When I first began these conversations about how UC Merced researchers are using AI, I expected to hear about tools, shortcuts, or clever workarounds. What I didn’t expect was how often the story turned out to be about thinking: how AI forces us to re-examine the logic of our own processes.
Suraj’s story captures that perfectly. He is a theorist who models life itself, using code and computation as his laboratory.Every one of these conversations ends up being about balance — curiosity versus caution, speed versus understanding. Suraj’s story reminds me that the most interesting part of AI in research isn’t the tool itself, but what it reveals about the researcher. To model life, you need both precision and humility; to use AI well, you need exactly the same.
Interested in participating? Email: sghosh@ucmerced.edu
