How a Bangladeshi scientist made it to MIT’s '35 Innovators Under 35' list
Mahi’s research attempts to solve one of robotics’ hardest problems: collecting enough real-world data to make robots useful in homes
Every year, Massachusetts Institute of Technology (MIT)'s renowned magazine, MIT Technology Review publishes a list titled '35 Innovators Under 35'. This year, one of those 35 was a Bangladeshi, Nur Muhammad Mahi Shafiullah.
At just 27, he was recognised for his pioneering work in robotics and machine learning. His inclusion placed him in the list of some of the world's most celebrated innovators, with past recipients including Elon Musk, Larry Page and Mark Zuckerberg.
"This recognition signals that my direction of research matters and deserves more time and focus. This kind of acknowledgement is definitely very encouraging," Mahi said.
Love for numbers
Growing up in Mirpur, Mahi's journey to the international stage was shaped early on by mathematics. He studied at Monipur High School and later at Dhaka College, where his talent for problem-solving quickly became apparent.
Between 2011 and 2014, he represented Bangladesh at the International Mathematical Olympiad (IMO), which led him to secure two bronze medals and a silver one. Besides IMO, he also competed at the International Olympiad in Informatics (IOI). These early achievements opened the door to one of the world's most prestigious academic institutions: MIT.
At MIT, he pursued his undergraduate studies in mathematics and computer science, later adding a Master's in electrical engineering and computer science.
"I began doing research in my third year at MIT," he recalled. That early exposure to academic research proved decisive for his later career trajectory. He then joined New York University's Courant Institute of Mathematical Sciences, where he completed his PhD in computer science this June. During his PhD, he had already moved from computer vision into robotics — a transition that shaped the next phase of his career.
Building smarter robots
Currently, Mahi is working as a postdoctoral researcher with Meta's Fundamental AI Research (FAIR) division, while also contributing to Berkeley AI Research (BAIR). His central research question is to explore how robots can function in the chaos of real human life.
"To explain it simply, imagine that we live in homes and apartments. That is one kind of world. Now imagine a factory or an industrial setting. That is another kind of world. There is a huge difference between the two. A factory environment is far more structured than a household one," he said.
"If you are manufacturing a car, the same screw may need to be fitted a thousand times a day, every single day. There is a high level of repeatability."
However, in case of tasks inside a home, each one is different. As Mahi said, "Suppose you want to make tea. The sugar may be in a different place each day. Even when boiling water, the way you do it will vary from one day to the next."
"Every problem and every solution will be different. Problems are highly individualised, and solutions must be tailored accordingly. You need to ask yourself: What are the challenges in my community, what are the local strengths, and how can I use those strengths to address the problems that matter most?"
This is where Mahi found his research interest. "There was this huge gap, and earlier there was no way for robotics to cross that gap," he explained. His PhD sought to bridge it.
His previous works included three major initiatives. Robot Utility Models focused on training machines to adapt quickly to new environments. The Dobb-E Project used inexpensive hardware to capture everyday household actions — opening cabinets, folding towels, tidying up — so that robots could learn from them.
Meanwhile, the OK-Robot Project aimed to teach robots to perform long-term tasks without retraining. Together, these projects demonstrated how deep learning methods, which had transformed natural language processing and computer vision, could also help robots operate in unstructured human settings.
Why it matters
The MIT Technology Review described his efforts as an attempt to solve one of robotics' hardest problems: collecting enough real-world data to make robots useful in homes.
Mahi and his collaborators devised creative solutions, such as mounting iPhones on sticks to record how people perform simple chores. These datasets were later used by Nvidia, Microsoft and Google. They also contributed to a dataset of 527 robot skills — such as rearranging kitchen objects or folding laundry — that allowed machines to tackle tasks they had never seen before.
The motivation for Mahi, however, lies beyond the technical challenges. "My focus is on how we can bring robots into the home, teach them to carry out household tasks, and train them to adapt to the preferences of a particular household," he said.
He imagines a future where robots act like helpers or butlers, that will ease the burden of repetitive chores. More importantly, he sees them as vital for an ageing global population.
As people live longer, family members often struggle to provide full-time care for the elderly. "Imagine if there were care-giving robots that could look after them," he suggested. "Then they would not have to move to an old age home, grow older happily, and enjoy life with their families."
He believes it is a direction that will define the next decade.
To young researchers
Thousands of students in Bangladesh are passionate about robotics. But unfortunately, opportunities are almost non-existent. Robots are expensive, universities lack facilities, skilled teachers are scarce — and whatnot.
However, acknowledging those shortcomings, Mahi believes that Bangladesh still has a major advantage: its diverse population. "People have different preferences and households, which is actually a rich source of data if you think creatively," he explained. Smartphones, for instance, can be powerful tools for gathering real-world data for robotics at scale.
His advice to young researchers is simple. Instead of chasing trends blindly, start with problems that feel personal and real. His own interest in home robotics stemmed from family concerns: living abroad while his parents stayed in Bangladesh made him think about how technology could ease the challenges of ageing and distance.
"It is about life — understanding people, the problems they face, and then thinking about how to solve them," he said.
He also emphasised that research should make use of local strengths. "Every problem and every solution will be different. Problems are highly individualised, and solutions must be tailored accordingly. You need to ask yourself: what are the challenges in my community, what are the local strengths, and how can I use those strengths to address the problems that matter most?"
