Are you a client?
Sign in to view the full news archive.
London HQ’ed startup Doubleword (previously known as TitanML), which provides a self-hosted inference platform for enterprises, has raised $12m in further funding. The round was led by Dawn Capital with participation from K5 Global, Hugging Face CEO Clément Delangue and Dataiku CEO Florian Douetteau.
Founded in 2023 by Meryem Arik, CEO, Jamie Dborin, CSO, and Fergus Finn, CTO, Doubleword has developed an enterprise AI model platform that enables one-click deployment of AI models across any environment, allowing teams to focus on building AI-powered products instead of managing DevOps. It is already used by firms such as AWS, Intel, Google Cloud, Nvidia and Meta.
In a blog post the company recently provided an update on the new name change, as well as how the business has changed since launch.
“Why ‘Doubleword’? In computing, a doubleword is a standard unit of data - but, for us it also captures a bigger shift. Words are the new interface to intelligence — but under the hood, it’s all bits, bytes, words, and doublewords. We take care of the technical complexity so our customers can access powerful AI with simple, natural language”
CEO Arik noted that when they founded TitanML, the AI world revolved around training - cost, infrastructure, breakthroughs. The company made a different bet, that inference, not training, would become the critical bottleneck to AI’s impact. Over the last few years this has taken the business from a small team of physicists solving tough research problems to a product-focused company helping enterprises simplify and scale self-hosted inference.
The Doubleword platform allows companies that are developing AI models to deploy on any hardware, with no vendor lock in. It allows one click deployment of any open-source, domain-specific, or custom Al model. The platform also acts as a unified control layer for model deployments, with a single pane of glass providing analytics into infrastructure and workload usage across cloud and on-premise environments.
It is a somewhat niche platform, given many companies are deploying AI models utilising platforms like OpenAI enterprise or off the shelf 3rd party software. That said the companies who are building and deploying AI models at scale, like the hyperscalers and those with a budget big enough to build from scratch, are surely going go find great value in a tool that can drive even minor efficiencies, with the cost of inference only going to be a rising issue for many organisations.
Posted by: Simon Baxter at 09:41