About
From memory research to practical AI.
HumemAI started from research into human-like memory systems for AI, then shifted toward building tools and products that make persistent, inspectable memory usable in real workflows.
Origin
The central idea is simple: conversations and data should not be treated as the same thing, and agentic systems need better memory structures than a single flat context window.
HumemAI was founded by Taewoon Kim, an AI researcher and engineer working on agentic memory: systems that help agents remember, organize, and reuse knowledge over time in ways that stay structured, inspectable, and practically useful.
Direction
Today, HumemAI sits between open-source infrastructure, applied research, and a hosted product direction. The goal is not to publish memory ideas in isolation, but to turn them into software that teams can actually use to build reliable agents.