Valohai

Run machine-learning training and deployment pipelines on whatever hardware you already have — AWS, Azure, OCI, OVHcloud, your own racks, or an air-gapped room — and keep a full record of which code, data and parameters produced which model. Their own framing is "the MLOps platform for ML Pioneers"; the four product pillars on the site are model workflows, LLM workflows (model comparison, context pipelines, regression testing, cost analysis), your infrastructure (multi-cloud, on-prem and hybrid, GPU efficiency, dataset caching) and lineage and reproducibility (automatic versioning, one-click reproduction, audit and compliance).