Last year, I started working on an AI that could locate images instantly. Nobody really believed in what I was doing. I was training models on city imagery, one city at a time. Miami. New York. Paris. London.
Each city took weeks. I kept improving the accuracy, kept adding more locations, kept pushing to see how far I could take it. But when I tried to explain what I was building, people just didn't get it. Maybe it was my environment. Maybe the people around me couldn't see what I saw. Or maybe I wasn't explaining it well. Either way, I spent a lot of time working alone, wondering if any of this would amount to anything.
Then a few people joined me. Not many. Just enough. And those few believers are the reason any of this exists today.
What we believe
We believe in building what should exist, not just what is easy. We believe that privacy and precision are not mutually exclusive. We believe that small teams with conviction can outperform large committees. And we believe that technology should be accessible, not locked away.
"We want to develop more machine learning systems. That's the honest answer: yes, we do. We're just still working out how."
— Internal Memo, 2025
The future
The question I keep getting is: "What will be the future of Oceanir? Is it only going to rely on just geo-estimation?" My answer is no. We want to build more. More models. More capabilities. Geo-estimation was where we started, but it's not where we have to stay.
The same techniques that power location recognition can be applied to scene understanding, visual verification, and things that haven't been tried yet.
Open research
We've opened a GitHub account and set up a Hugging Face presence. The infrastructure for sharing our work is there. We're exploring ways to share our work with the research community.
To the few who believed when no one else did: I haven't forgotten. I never will. We're just getting started.
