07 / Selected work
AI product · System architecture
Multi Memory
A graph-based AI memory system for persistent context in long-running LLM workflows.
05documented product views
07connected system capabilities
01persistent context layer
24/7context available across chats
Project overview
Separate conversations become one connected memory system, so useful context can persist instead of disappearing between sessions.
01 / Problem
Context disappeared between chats
Long-running AI work was split across isolated message threads. Important decisions, recurring topics and reusable knowledge had to be repeated or manually reconstructed every time.
02 / Solution
Build memory as a graph
I designed a fullstack system combining chat history, automatic memory extraction, context-aware retrieval, tagging, used-context visibility and graph-based exploration in one product interface.
03 / Outcome
One navigable context layer
The prototype connects conversations, projects and recurring ideas as persistent memories. Relevant context can be retrieved for new work while remaining visible and inspectable to the user.