Belvedir
Belvedir is a YC Summer 2026 (S26) company. Based in San Francisco, CA, USA. Team of 1.
About Belvedir
Summary
Belvedir builds a custom model factory for teams running AI agents. Its SDK captures agent traces, tool calls, and replies, then automatically trains task-specific models, evaluation harnesses, and routing logic, deploys them privately, and retrains them on live usage. It replaces the manual work of dataset curation, fine-tuning pipelines, and inference optimization that today keeps custom models out of reach for most builders.
Business model
$npm install belvedir suggests a developer SDK entry point; pricing structure not public.
Go-to-market
Targets developers and teams building AI agents, with a self-serve SDK install as the starting point.
From the Y Combinator listing
The easiest way to make private AI models.
Private AI models are going to win. Belvedir is the infrastructure needed to make trillions of them. Belvedir makes it cheap and easy for anyone to make private AI. Connect us to your agent traces through our SDK and we autonomously train custom models + harnesses + routers, privately deploy them, and recursively improve them as they are used. Our customers have seen up to 10x reductions in inference costs and 2x increases on benchmark scores.
Facts
- B2B / B2B
- San Francisco, CA, USA
- 1
- Active
- Early
Founders (1)
Score every batch against your own thesis
GPAgent's sourcing agent reads every company in a batch, scores each one against your fund's thesis in a private workspace, and lines up the founders worth meeting before Demo Day. You decide who to contact. Your scores, notes, and pipeline never appear on these pages.
How Accelerator intelligence works