Valohai

Valohai is a Finnish MLOps platform that runs machine learning workloads on infrastructure the customer already owns. It does not sell its own compute. Teams define jobs in containers and launch them from a CLI (vh), an API or a web UI. Valohai then provisions and auto-scales machines on AWS, Azure, Google Cloud, Oracle Cloud, OVHcloud, Scaleway, on-premises hardware or air-gapped environments. Every experiment, dataset and model is versioned automatically, which gives full lineage and reproducibility. Pipelines can include conditional quality gates, parallel runs and human approvals, and a model registry handles approval workflows. Newer features cover LLM work, including model comparison, regression testing and cost analysis, with Langfuse tracing and agent skills for coding assistants such as Claude Code and Cursor. Distributed training supports PyTorch Distributed, DeepSpeed, Horovod and Accelerate, and trained models can be served as HTTP endpoints on Kubernetes. Valohai is aimed at data science and ML engineering teams in regulated or security-conscious enterprises that need to keep their data and compute in their own environment. The company reports a SOC 2 Type II audit. Pricing is a fixed fee per user per month and includes unlimited projects, experiments, pipelines and deployments. Prices are not published and require a custom quote, and a 14-day free trial is available. The main limitation is that customers still pay their cloud or hardware provider for compute. Teams also need some setup effort to connect Valohai to their existing environment, and buyers cannot compare costs without talking to sales.