Meet Adam Munawar Rahman: Leveraging AI for Humanitarian Use

Adam Munawar Rahman, an MS student in Computer Engineering at NYU Tandon, recently took second place in the 2025 ARM AI Developer Challenge, competing against more than 1,600 participants worldwide.

His project, DreamMeridian, runs AI-powered spatial queries entirely offline on a simple Raspberry Pi, a project designed for refugee camps, disaster zones, and other humanitarian scenarios where internet connectivity is limited or nonexistent.

We caught up with him to learn more.

What drew you to NYU Tandon for your graduate work in Computer Engineering?

I'm a software engineer at IBM, where I work on mainframe platforms, and the company has been very supportive of my desire to work and study concurrently. What really drew me specifically to computer engineering was the opportunity to bridge the gap between hardware and software. I've been particularly interested in low-power devices like the Raspberry Pi, which can operate in low-connectivity scenarios where traditional cloud-based solutions just won't work.

Can you explain what DreamMeridian does in plain language?

DreamMeridian is a geospatial AI tool that provides real-time routing and resource location information on low-power devices without relying on cloud infrastructure. Think about being in a refugee camp or disaster zone where you need to find the nearest medical tent or water source. You can ask the system in natural language—just like you'd ask a person—and it gives you directions and information, all running locally on a device about the size of a deck of cards.

The key innovation is that I optimized the AI component for natural language understanding and decision-making, while delegating computationally intensive tasks—such as routing calculations—to traditional algorithms. That's what makes it efficient enough to run on resource-constrained hardware.

What's next for the project?

I'm planning to collaborate with community organizers here in Brooklyn to explore practical use cases.

I recently took part in a climate resilience hackathon called “Urban Futures” at the American Museum of Natural History, and my team won Best Community Communication Tool. We worked directly with organizers from the Pitkin Ave. BID and the Ocean Hill-Brownsville Neighborhood Improvement Association to build a climate-aware emergency routing tool for Brownsville, Brooklyn: one of New York City's most climate-vulnerable and underserved neighborhoods. Their feedback ("We've been studied enough") shaped our approach: building tools their existing planners could actually use. That experience reinforced my interest in community-centered design for DreamMeridian.

What was the ARM AI Developer Challenge like?

It was incredibly competitive. When you're up against 1,652 participants from around the world, the bar is high. What made me decide to enter was that it aligned perfectly with what I've been working toward—exploring practical applications of AI in humanitarian scenarios and pushing the boundaries of what's possible on edge devices.

This extends work from a previous UNICEF Innovation internship you completed. How did that experience shape DreamMeridian?

That UNICEF internship was really formative. I was writing Python data pipelines to calculate distances between schools and health facilities across UNICEF programme countries, using geospatial algorithms and OpenStreetMap road networks. It got me thinking: how do we bring powerful capabilities to places where the infrastructure we take for granted simply doesn't exist? That algorithmic work is what DreamMeridian extends: same spatial reasoning, now running offline on edge hardware.

I'd love, if possible, to reconnect with UNICEF to see if there are opportunities to deploy this in various humanitarian contexts. The goal has always been to move this from concept to actual implementation, where it can make a difference.

My interest in humanitarian work actually goes back even further, to high school, where I worked on a computational research project: "Jails by Java," which used agent-based modeling to simulate racial disparities in incarceration rates based on Bureau of Justice Statistics data. That computational social science research helped shape my focus.

Why is offline capability so critical for humanitarian use cases?

In a disaster zone or refugee camp, you can't count on internet connectivity. But people still need to navigate their environment, find resources, and make informed decisions. When connectivity is limited or nonexistent, having AI capabilities that run entirely locally can be the difference between finding help quickly or not finding it at all.

You've been at Tandon for only one semester so far. Has the program already influenced your work?

The experience has been really positive. Graduate programs like the one I’m in at Tandon are valuable for deepening technical understanding and fostering innovation in ways that purely industrial work sometimes can't. The interdisciplinary environment here has been particularly valuable—I'm exploring connections between urban planning and AI, which opens up entirely new ways of thinking about these problems.

What excites you most about the intersection of AI, edge computing, and social impact?

The democratization of technology. We're at a point where powerful AI capabilities can run on affordable, accessible hardware. That means communities that have been underserved by technology can now benefit from cutting-edge tools without requiring expensive infrastructure. Edge computing and AI together have enormous potential to address real-world problems in ways that were simply impossible just a few years ago.

DreamMeridian is open source and available on GitHub. Learn more on DevPost or watch the demo video.