AI-prototyping

explorations

PERSONAL PROJECTS

2025/ 2026

MILECRAFT.

I really enjoy getting deeper into points and miles to get better flight experiences. But one recurring pain point is figuring out the best stopovers and routes, and their costs in points. Most existing tools display flight segments separately, making it hard to visualize combinations. And the best deals usually depend on knowing your stopover options before you even start searching.

So I decided to build the prototype "Milecraft" to explore this combination: possible stopovers vs. price in points. I used SeatsAero to find flight segments and connected to their API to get real flight data. Then I used Cursor to build the experience, with an interface very close to what I had in mind.

The outcome is a website that gives a better sense of possible stopovers, routes, and costs in points. The map simplifies the search, making it easier to understand distances, directions, and the number of flights.

It was a really fun process to take a complex personal pain point and explore how I could build something more intuitive and usable.

PLOC - VERSION 2

Continuing my previous study, this time I wanted to go beyond a 2D screen and play with the features already available in our devices to enable different interactions.

For this phase, I wanted to experiment with the gyroscope on a tablet. Players can now tilt the device to slide an ice block and freeze the lava, creating a path for the character to reach the diamond. It’s a simple mechanic, but it adds a playful interaction beyond taps and swipes.

It makes it so much easier to exchange ideas, discuss details, and get feedback.

PLOC - VERSION 1

I wanted to get more hands-on with my learning at MIT xPRO course “Designing and Building AI Products and Services.”. So I revisited a game project from my design degree to explore AI tools in practice. The game was designed to encourage logical reasoning in kids, where players select command sequences to move the character and reach the diamond in each level. Back then, I didn’t have the knowledge to prototype it for real, but this time I wanted to see how I could use AI tools to push my design process further.

As part of the exploration, I tested different AI tools. I used Notion AI and ChatGPT to support my early design phases, helping me organize ideas, write clearer prompts, and plan next steps. Then I used Cursor to actually build and test a prototype. This video shows the webapp prototype I built with the first two levels working.

I had to learn not only how to refine my prompts, but also how to communicate better, providing clearer context and guiding the AI to achieve the behaviors, UI, and animations I was envisioning. Many times, I had to completely rethink my approach or use AI to help me understand the code structure so I could fix it myself and get closer to the results I wanted.