Paras Stefanopoulos

Post-Training Team

Model performance

Optimizing delta weight syncs for managed rollouts

Paras Stefanopoulos

The trainer publishes each checkpoint or delta once to Amazon S3 and the Baseten Delivery Network distributes it to independent rollout clusters without a direct trainer-to-GPU connection.

Foundations

Open-source LLM training is a mess. Here is how it all works.

Paras Stefanopoulos

AI engineering

OpenClaw + Kimi K2.5 on Baseten = frontier agent performance with OSS

Alex Ker

Alex Ker

1 other

openclaw + kimi k2.5

Explore Baseten today

Start deploying Talk to an engineer