PhotoSage, a GenAI innovation that makes photo retrieval more convenient
Your motivation behind building this.
The idea for PhotoSage came from a common frustration: scrolling endlessly through a phone to find the needed photos. We found this process stressful and tedious, and we realised we weren’t the only one facing this problem. Almost everyone struggles to locate specific images in a sea of media. While existing image search tools did prove helpful, they fell short of fully solving the problem — this sparked the idea of creating something better. This motivation grew stronger after we took a trip to Mysore with our friends.
We took around 3,000 photos during the trip, and finding specific pictures was incredibly time-consuming when we tried to post them on social media — like photos near a particular church, temple, or dosa shop. That’s when we decided to build PhotoSage to make photo retrieval more convenient for users like us.
What makes PhotoSage different from other photo search apps such as Google Photos?
We wanted PhotoSage to be faster and more precise than existing photo search apps. Google Photos relies on cloud-based AI, which means images are processed on remote servers. We took a different approach, running everything on-device to keep searches private and instantaneous. Speed was another priority.
Using an embedding-based algorithm and Apple’s ML APIs, we ensured PhotoSage could retrieve images faster than both Google Photos and Apple’s on-device search.
Accuracy mattered just as much. Instead of matching simple keywords, our AI understands detailed descriptions like “a dog playing in the garden” or “a sunset with a red car.”
We also didn’t want users scrolling through irrelevant images to find what they needed, so we made it a priority to minimize false positives. With an Android version in development, we’re refining it further.
