How they use Nebius AI Cloud

Seqera has run Nextflow protein-design and structure-prediction pipelines on Nebius AI Cloud. Its team describes testing a pipeline on an H100 GPU virtual machine and running a protein-binder design project on managed Kubernetes. The company also publishes cluster and workflow configuration for the joint work. For the January 2026 Nipah binder project, the production run scaled to 16 H100 GPUs for parallel Boltz2 validation. CPU stages handled ranking, ProteinMPNN sequence redesign and analysis. Seqera reports completing the workflow of 2,000 initial designs and 200 final candidates in a few hours.

In an earlier Chai-1 demonstration, Seqera predicted structures for 84 sequences using 24 H100 GPUs through Slurm and Soperator. Seqera identifies the example as nf-chai predictions for CASP15. The demonstration used single-sequence mode without multiple-sequence alignment and produced 420 structure predictions. Seqera used Apptainer to share container images between nodes and pre-downloaded Chai-1 reference weights for reuse across workflow tasks. A separate presentation describes moving Nextflow jobs to Nebius and reports a fourfold improvement for the deployment shown. In August 2026, Nebius reported a fourfold speedup for Seqera’s protein structure prediction through NVIDIA NIM compared with launching containers natively. These figures refer to particular demonstrations. Seqera also works with Nebius as a technology partner.

For its Kubernetes deployment, Seqera used Terraform to configure an autoscaling cluster and Nebius-managed GPU drivers, device plugins and Grafana dashboards. Nebius describes a jointly engineered Seqera integration supporting Nextflow and nf-core pipelines on its GPU clusters with one-click Terraform deployment. In March 2026, CEO Evan Floden reported success running Nextflow with Nebius and praised its GPU availability and interface.