Inferact

Inferact builds AI inference infrastructure with a single, precise objective: to make serving AI as simple as spinning up a serverless database. The company works to close the gap between capable AI models and the systems that serve them, driving down the cost and latency of inference so that access to powerful models is not confined to those able to build bespoke infrastructure.

The firm's technical lineage is its distinguishing feature. Its founders created and maintain vLLM, the most widely used open-source LLM inference engine. The project counts more than 2,000 contributors, supports over 500 model architectures and more than 200 accelerator types, and runs at global scale. Work spans LLM inference, inference optimization, accelerator optimization, model serving, mixture-of-experts models, multimodal AI systems and agentic AI.

Inferact operates at the intersection of open-source software, AI/ML infrastructure and cloud computing, with a global footprint. Engineers joining the company work on problems of genuine industry consequence, in a team whose open-source output already underpins much of the world's model serving, and whose stated mission is to accelerate AI progress for the industry as a whole.