Memori
The personal memory profile for AI. Keep your projects, notes, and meetings in one memory you own, and let every AI you use — ChatGPT, Claude, Gemini — work on what matters to you.
User-owned memoryAI contextMemori
CEO & Cofounder, Memori
Fernando
Torres
I’m building Memori, the personal memory profile for AI.
Stanford GSB MSx ’26 · Previously Business Development Director at Chinoin (pharma, 35+ countries)
What I work on
One thesis, three threads
The memory behind your AI should belong to you. Everything here is a way of building, proving, or explaining that.
Featured work
Projects
Memori, comicMemori, and the Stanford GSB research they grew out of.
Featured research · Stanford GSB
Why 90% of AI agent pilots never reach production
What Fernando Torres and Shekhar Bhende found at Stanford GSB about why enterprise AI agent deployments fail, and what to build instead.
90%
Cited pilot failure rate
Enterprise AI agent pilots that never move past proof-of-concept.
30–40%
Model contribution
Share of production success interviewees attributed to the model itself. Infrastructure is the rest.
92%
Sources mentioning integration
System integration appeared in 92% of sources. It was not the primary blocker.
Latest writing
From the blog
Research essays on production AI agents, comicMemori episodes, and field notes.
ThemecomicMemori
Vector databases, the map drawer at scale
The map drawer expands into several index mechanisms rather than one magical set of partitions.
ThemecomicMemori
Cosine similarity, comparing directions
Ria draws two nonzero vectors from a common origin and rotates one through aligned, perpendicular, and opposite positions while Ava watches the score.
ThemecomicMemori
Similarity search, finding neighbors on the meaning map
Ria places a query card into the map drawer. A flat exact baseline measures every stored vector.
Get in touch
Working on memory, agents, or context?
I read everything. Questions about Memori, the research, or comicMemori are all welcome.