beta9

Why we included this project

Beta9 turns GPU-backed AI work into plain Python functions. Mark a function with a decorator and it can run as an autoscaling inference endpoint, an isolated sandbox, or a background job, with cold starts under a second and scale-to-zero by default. That covers LLM inference, fine-tuning, and batch processing on H100s and 4090s without assembling a scheduler, container runtime, and queue by hand. You can run it on Beam's managed cloud or self-host everything and keep control of your GPUs. It suits teams that want to ship model workloads without operating the infrastructure underneath.