Boltz-2 Inference on Nebius MK8s with GPU and Shared Filesystem

This repository contains scripts and Kubernetes manifests for running Boltz-2 — an open-source biomolecular foundation model — on Nebius AI Cloud using a GPU-enabled Kubernetes (MK8s) cluster and a shared filesystem.

Boltz-2 predicts both 3D protein–ligand complex structures and binding affinities, enabling fast in silico screening for drug discovery.

Check the full, step-by-step tutorial here

---

Features

• Automated setup of a Nebius MK8s cluster with GPU nodes.
• Shared filesystem (PVC) for storing input and output data.
• Inference with parallel GPU jobs.
• Scripts for uploading input data and downloading results from PVC.