Moon AI
Moon closes Series B, led by Sequoia and General Catalyst →

Own your AI.
Own your models.

Build, own, and deploy specialised models trained on your work. Drive business value through reinforcement learning, in your cloud.

Book a demo
View all use cases
Use cases
All Active Archived
SupportRL-24

Auto triage every inbound ticket, by tone.

Reads the message and routes on urgency plus product surface. Two week RL loop from CSAT.

Accuracy94.2%
SalesRL-08

Draft the follow-up in the rep's actual voice.

Ingests the last 300 emails per rep. Learns cadence, closes, and the way each one signs off.

Open rate+38%
FraudRL-17

Score every transaction, without a rule engine.

Continuous online learning from analyst labels. Beats the vendor tool by 6 basis points on false-positives.

False-positives-42%
01 · What Moon is

A closed-loop stack for specialised models.

Three surfaces, one runtime. Moon gives your team the training loop, the deployment layer, and the evaluation harness in one place, no vendor lock-in.

i.

Training loop, in your cloud.

Bring your own weights, your own labels, your own retention policy. Moon never leaves the perimeter you define.

ii.

Deployment, without the airlift.

Ship the model to a serving endpoint in one click. Blue-green, canary, and shadow deploys, out of the box.

iii.

Evaluation, that reads the room.

Twelve built-in harnesses plus your own. Score every model against every task, every night, automatically.

What you get, versus what you pay for elsewhere.

Moon ships the closed loop. Everyone else ships a piece of it and calls the rest a partnership.

LegacyMoon
Own the weightsOn paperBy default
Train in your cloudAdd-onYes
Reinforcement loopDIYBuilt in
Eval harnessSeparate toolYes
DeploymentTwo vendorsOne runtime
Time to first prodNine weeksFour days
Moon is the first platform where owning our models is not a slide. It is a Tuesday.
Marina Volkov · CTO, Datadog

Bring your data.
Keep your model.

Book a demo Read the technical brief