Core Product
Lyceum is one platform with three product surfaces, all built on the same core (Gateway → Streamer → Hydra-provisioned compute) but exposed to customers in different ways:
-
Execlets — Serverless
Per-job execution. Customers submit code or a Docker image; we schedule it on a matching GPU, run it, and stream the output back.
-
Bare VMs via Hydra — Direct access
A whole GPU machine on request. Hydra provisions it with a compute provider and hands over SSH access to the customer.
-
Inference via Iris — Managed serving
Deployed, auto-scaling model endpoints with an OpenAI-compatible API.
Overhaul
The job scheduling of the streamer is currently getting overhauled by building our platform on top of Kubernetes. The streamer with its execlets can therefore be considered as legacy and should not consume much of our work.