1st Place with AI solution @ DMZ Japan & YCombinator Hackathon

This project was developed across two separate hackathons: - DMZ Japan x JVA Hackathon, where the original concept was created - Y Combinator Hackathon, where the idea was expanded, refined, and improved Using my academic background in environment and development, I shaped the concept around pressing sustainability challenges in Japan—specifically clean energy adoption and resilient infrastructure.

Back to projects

Problem

Japan faces two urgent national priorities:

Accelerating clean energy adoption
Strengthening disaster resilience

Yet assessing where solar panels can be installed is slow, manual, and costly.
Energy providers must check rooftops one by one delaying the shift to renewables and limiting preparedness.

Our Idea

Our Goal

Create a rapid, scalable solution aligned with SDG 7: Affordable & Clean Energy by using technology to identify optimal solar installation sites across Japan.

Our Solution

We built a platform that uses satellite imagery + an AI/ML rooftop classification model to instantly pre-qualify roofs suitable for solar. This allows governments, energy companies, and NGOs to quickly assess opportunities at any scale—city-wide to region-wide.

Process

Because this hackathon required speed and clarity, we created a focused, high-intensity process:

Define the problem & user needs

Research energy + disaster landscape in Japan
Collect satellite imagery datasets
Build and train an AI/ML model to classify rooftops
Design the platform (flows, dashboards, filters)
Validate feasibility with rapid testing
Deliver pitch supported by a working prototype

This structured flow allowed us to move fast while validating decisions at every step.

Research

We explored three areas to ground our solution:

User Flow

Before designing the interface, I mapped the essential user flow:

Wireframe

I converted the flows into wireframes and then polished UI screens focusing on:

Clean, map-first layout
Color-coded rooftop classifications
Accessible sidebar filters
Actionable insights summary (e.g., “1,247 solar-ready rooftops identified”)
Export & reporting tools

These designs made complex AI output understandable at a glance.

ML Solution

We evaluated three imagery types:

Commercial High-Res
- Extremely accurate
  – Expensive, strict licensing
Japan Government Imagery (GSI)
- Free/affordable, good coverage (20–40 cm)
  – Quality varies
Public Satellite
- Free
  – Too low resolution for rooftops

Final Approach:
Use free GSI orthophotos for training, and paid high-res imagery (Maxar/AW3D) only for priority zones to balance cost and accuracy.

AI Solution

Using the wireframes, userflows and research we made a working prototype using Loveable.

Result

Our solution won the DMZ Japan x JVA Hackathon, earning recognition for its practical impact on clean energy adoption and disaster resilience. We then advanced the idea further for the Y Combinator Hackathon, where we received the “Most Venture-Backable” Award, highlighting the concept’s scalability and market potential.

These wins were driven by our ability to think beyond traditional energy models, apply creative AI/ML solutions, and ground the concept in my personal commitment to sustainability and environmental development.

Next Steps

Our team was accepted into DMZ Japan, a program that supports early-stage startups through mentorship, technical guidance, and market validation. As part of this Basecamp phase, we plan to continue refining the model’s accuracy, expanding dataset coverage, strengthening the platform’s UX, and validating real-world use cases with energy partners. This next stage will help us fine-tune the concept into a viable, scalable solution for Japan’s clean energy and resilience needs.

Learnings

This experience reinforced how much I enjoy contributing bold ideas and shaping solutions that tie back to real-world impact. It also taught me the value of clear communication and collaboration working smoothly as a team was essential to building and pitching a complex AI concept under tight deadlines. Moving forward, I’m motivated to bring this same creativity, teamwork, and purpose-driven thinking into my professional work.

More projects

Check out some of my favorite & most recent projects.

Fit-sizing UI solution with a 4.7% increase in widget clicks

In-house Product Designer