MemoryOS

Turn your screenshots, photos, and docs into an instantly searchable digital memory. Just ask MemoryOS, "Where was that Tokyo hotel?", and get immediate, accurate answers from your own saved files.

Comment

Inspiration

We all take dozens of screenshots, save photos, download PDFs, and bookmark receipts every week, only to completely forget where they are when we actually need them. Traditional search only works if you remember exact file names or tags. I was inspired to build MemoryOS to give everyone an "extended digital memory", an AI assistant that instantly connects the dots across all your saved visual and textual information so you never lose track of important details again.

What it does

MemoryOS turns your chaotic collection of screenshots, photos, receipts, tickets, and documents into an instantly searchable, interactive personal memory vault. Instead of digging through endless photo albums or folders, users can simply ask natural questions like "Where was that Tokyo hotel my friend recommended?" or "How much did I pay for my last flight?" MemoryOS analyzes the content, extracts relevant context via OCR and multimodal reasoning, and delivers exact, verified answers instantly.

How i built it

  • Mobile App: Built for cross-platform performance and a seamless capture-to-search user experience.
  • AI & Vision Pipeline: Powered by advanced multimodal models and OCR to process images, parse unstructured text, and build semantic contextual embeddings for accurate document reasoning.

Challenges we ran into

Structuring and cross-referencing completely unstructured visual data (like low-res receipts or crowded screenshots) posed a significant challenge. Ensuring high-accuracy semantic retrieval without high latency required fine-tuning our prompt pipeline and optimizing local/cloud hybrid data extraction.

Accomplishments that i'm proud of

  • Delivering accurate, conversational answers across hundreds of mixed screenshots and files in under two seconds.
  • Implementing a complete, end-to-end monetization and retention workflow using RevenueCat.
  • Creating an intuitive "Ask My Memory" interface that feels like talking to a personal assistant who has perfect recall.

What i learned

I learned how to effectively design AI-native utility applications around real-world user habits, balance complex AI reasoning latency with fast client UI updates, and build scalable subscription architecture for mobile apps.

What's next for MemoryOS

  • Timeline Intelligence: Automatically generating chronological trip/event recaps from imported media.
  • Privacy-Safe Shared Memories: Allowing users to curate and share specific memory collections (e.g., "My Favorite Tokyo Spots") without exposing private data.
  • On-Device Search: Bringing core indexing local for offline-first privacy and speed.

Built With

Share this project:

Updates

Submission history