SofiaLLM Ops
Institutional LLM operations
Fine-tune, serve and measure language models on your own HPC.
Temporal inference on demand
Bring up any base or fine-tuned model on 1–4 HPC GPUs and hold them only for the minutes you need.
Continuous endpoints
Always-on, OpenAI-compatible endpoints for the models your group runs in production.
Fine-tune and prove it
LoRA and QLoRA fine-tuning jobs, with evaluation runs and side-by-side comparison built in.
Universidad de Salamanca
Project SOFIA · CDTI–AEI
Sign in
Use the institutional Google account linked to your research group.
Continue with Google
Single sign-on through your university identity provider.
API keys, GPU budgets and token quotas are attached to your account.
Usage is recorded per key, project and model.
Access is limited to enabled accounts. Contact a platform administrator if your group is not yet provisioned.
Terms of use · Data processing