Memversa

• 

  People. Places. Stories.

Inspiration

Photos preserve what a moment looked like, but not necessarily what actually happened. Years later, friends often remember the same night differently — one person remembers a place, another remembers a conversation, while someone else has a photo that changes the whole story.

Memversa turns those fragments into a shared investigation of the past.

What it does

Memversa finds moments from your photo library and creates a shared Memory Room for the people who were there.

Friends contribute memories, photos, and missing details. AI connects these fragments, finds contradictions and unanswered questions, and turns them into mysteries to investigate together.

Instead of simply generating a summary, Memversa helps answer questions like:

• Where did we go after this?

• Who was actually there?

• What happened between these two photos?

• Why do we remember this differently?

Photos and metadata become evidence, while each person's memory becomes another perspective.

Once an event is reconstructed, Memversa can go one step further with What If? — exploring alternative paths based on moments and decisions from the past.

The long-term vision is a Second Life built from your real memories: an explorable network of people, places, stories, decisions, and the lives that could have unfolded differently.

How we built it

Memversa is an iOS app built with SwiftUI and PhotoKit. Photo discovery happens on-device, while selected memories can be shared through persistent Memory Rooms.

The backend is built with FastAPI on Google Cloud Run, with Firestore for persistent rooms and collaborative memory data, Cloud Storage for photos, and Gemini/Vertex AI for memory analysis and reconstruction.

AI analyzes memories and evidence to identify missing information, conflicting accounts, and potential mysteries without treating generated guesses as facts.

Challenges we ran into

One of the biggest challenges was separating reconstruction from invention. Human memories are incomplete and often contradictory, so the AI needs to distinguish between evidence, individual recollections, uncertainty, and generated possibilities.

We also designed the system so that selecting the same photo event again resolves to the existing Memory Room rather than creating duplicate memories.

Accomplishments that we're proud of

We built a working collaborative memory system where multiple people can reconstruct the same event together. Memversa can identify missing details and conflicting memories, connect them with photo evidence, and turn them into mysteries that friends can investigate.

What we learned

We learned that memory is rarely a single version of the truth. The most interesting moments appear when different people remember the same event differently. AI is most useful here not as a storyteller, but as an investigator that connects evidence, asks the right questions, and preserves uncertainty.

What's next for Memversa

We want to connect reconstructed memories across a person’s lifetime into a Second Life — a network of people, places, relationships, and decisions. The next step is “What If?”, allowing users to explore how their lives might have unfolded if key moments had gone differently.

Built With

• cloud-firestore
• fastapi
• gemini
• google-cloud
• google-cloud-run
• photokit
• python
• swiftui
• vertex-ai