
Hey everyone, we are Taerim and Tiantian, and we are building Nines.
https://www.youtube.com/watch?v=b9PAOiSU7ek
AI data centers are being built faster than they can be staffed, and in places where technicians don’t live. Every one of them still depends on a person walking to a rack. That costs money in two ways:
What we’re building:
On new buildouts, our robots handle the commissioning work, like patching fiber across thousands of ports. On live sites, they handle the maintenance that costs operators money when a node sits idle: swapping drives, reseating optics, restarting servers.
@Tiantian Li and I believe compute is going to places humans can’t or won’t go, and the data centers there will have to run themselves. We’re building toward that one piece at a time.
Today, AI models only optimize themselves in software. Soon, data center orchestration software will detect when a training run needs a new network layout or more storage, and send the physical work straight to our robot fleet.
We're building the automation layer for server rooms: a fully software-defined, physically adaptive AI factory where AI can manage and optimize its own hardware. Fiber patching is step one, and we're building the rest.
If you build or operate AI data centers, we’d love to chat. Or, if you know people at data center operators, we’d appreciate an intro.
Who we are:
We have spent our careers on hard physical and scaling problems. @Tiantian ran large-scale inference at Meta, and watched data center construction and maintenance become the limit on compute. @Taerim has spent her career shipping 0-to-1 systems, including taking Patreon’s livestream from the first line of code to millions of users.
Follow along on X for our updates: https://x.com/ninesrobotics @ninesrobotics