Geniusrise

Geniusrise is a modular, loosely-coupled framework for operationalizing AI models as microservices. It exposes REST endpoints and a CLI for hosting inference on open-source or closed-source models, running bulk inference jobs, fine-tuning, and wiring together multi-model pipelines. It ships runners for Kubernetes, Docker, Docker Swarm, and Apache Airflow, so the same pipeline can run on a laptop or a production cluster.

It is aimed at ML engineers and platform teams who want an opinionated scaffold for shipping inference services without gluing together FastAPI, Hugging Face, queue runners, and storage adapters by hand. The framework includes 40+ data-source connectors (databases and streaming), document preprocessing, and OCR, which makes it more practical than a bare model server for real ingestion-heavy workloads. The project is open source and self-hosted; there is no SaaS tier or published pricing.

Because it leans on the underlying model ecosystem (transformers, audio/vision models, optional closed APIs), Geniusrise functions as a control plane rather than a model provider. The trade-off is the usual self-hosted reality: you own the infra, the GPU bill, and the upgrade path.