As the generative artificial intelligence phenomenon gathers pace, many enterprises are still working out how to get the best out of the large language models on which GenAI is built. One big dilemma is whether to go for a proprietary LLM, requiring a license and subject to strict rules about how it can be used, or an open-source model that the enterprise is free to customize for its own purposes.
San Francisco-based Pipeshift believes a growing numbers of companies will opt for the latter option, given the freedom and control it offers – but only if they can find a way to deploy open-source LLMs quickly and effectively within their businesses. The start-up will today announce it has raised $2.5 million of seed finance as it rolls out an infrastructure platform it claims will make it much easier for enterprises to do exactly that.
“We think 2025 will be the year when GenAI transitions into production, with engineering teams witnessing the benefits of using open-source models in-house,” says Arko Chattopadhyay, co-founder and CEO of Pipeshift. “This offers high levels of privacy and control alongside enhanced performance and lower costs – but it is also a complex and expensive process requiring enterprises to stitch multiple components together.”
Many enterprises lack the resources to do that work – the time, money and expertise. Pipeshift’s big idea is a platform that operates as a one-stop-shop for this orchestration process. It provides pre-built capabilities that enable enterprises to train, deploy and scale open-source LLMs within their businesses, so that small armies of engineers don’t have to spend weeks and months doing the heavy lifting normally required.
Such an approach not only provides convenience but also offers flexibility, Chattopadhyay argues. “We’re two years into what will probably be a 10-year development cycle for GenAI, so why would you tie yourself into a single LLM at this stage,” he says. “Our platform makes it much easier to change models or to stitch LLMs together as you see fit.”
Chattopadhyay met his co-founders Enrique Ferrao and Pranav Reddy when all three were undergraduates working on a defense robotics project supported by Nvidia, Dassault Systems and SICK Sensor Intelligence. The trio launched Pipeshift last year, initially working with around 30 customers in beta. Around a fifth of that cohort has already stepped up to become fully-fledged clients as Pipeshift continues to develop commercially.
At one of those customers, NetApp, director of software engineering Anu Mangaly points to the platform’s practicality and cost-effectiveness. “Pipeshift’s ability to orchestrate existing GPUs is extremely impressive, allowing businesses reduce their compute footprint and costs in production, while delivering enhanced user experiences that are also private and secure,” she says.
Pipeshift’s investors are also convinced about the company’s potential. Today’s fundraising is led by Y Combinator and SenseAI Ventures, with additional participation from Arka Venture Labs, Good News Ventures, Nivesha Ventures, Astir VC, GradCapital, and MyAsiaVC. The round has also attracted a number of prominent business angels from the tech sector.
“Enterprises prefer open-source GenAI for the benefits of privacy, model ownership and lower costs but transitioning GenAI to production remains a complex and expensive process requiring multiple components to be stitched,” says Rahul Agarwalla, managing partner at SenseAI Ventures.
“Pipeshift’s enterprise-grade orchestration platform eliminates the need for such extensive engineering investments by not only simplifying deployment but also maximizing the production throughput.”
Pipeshift will use the cash raised to perfect the platform – Chattopadhyay is confident of completing upgrades in the coming months, though he says it already slashes time-to-deploy – and to build profile for the business. “We think we can democratise access to these open-source models, but business leaders need to be made aware of what’s possible, rather simply accepting the models that their existing account executives want to sell them.”
Read the full article here
