While a growing number of companies are investing in AI, many are struggling to bring AI-based projects into production – much less generate a meaningful ROI.
The challenges are numerous. But a recurring problem is data management. The data businesses need to train, run, and refine AI models is disorganized, siled, and otherwise unoptimized. In a 2022 survey by Great Expectations, an open source data benchmarking platform, 77% of organizations said they were concerned about the quality of their data.
Startups that promise to solve these data problems are raising money.
On Wednesday, Weka, a data pipeline building platform managing a range of data sources, types and sizes, announced it has raised $140 million in a two-part Series E round ( $100 million and $40 million) led by Valor Equity Partners. , with participation from Nvidia, Norwest Venture Partners, Micron Ventures, Qualcomm Ventures, Hitachi Ventures and others. The oversubscribed funding round values Weka at $1.6 billion after money, double the company’s previous valuation.
Weka CEO Liran Zvibel and fellow Weka co-founders Maor Ben-Dayan and Omri Palmon met during the founding of data storage startup XIV, which IBM acquired for $350 million in 2007 The trio remained at IBM for several years, but eventually left to pursue other independent projects.
However, the data management problem continues to plague Zvibel, he says.
“I was frustrated and disappointed to see customers forced to use disparate, siled data infrastructure solutions that were unnecessary, expensive and complex to deploy, manage and maintain,” he said. “The problem has become particularly evident with the rise of cloud computing and the advent of high-performance computing, machine learning and early AI workloads. »
So in 2013, Zvibel recruited Ben-Dayan and Palmon to create a new set of data tools, one that could provide a better approach to storing, managing, and moving data.
“We envisioned a platform powerful enough to support the performance demands of next-generation computing hardware and large-scale, data-intensive workloads in demanding, distributed environments,” Zvibel. “To meet the needs of modern workloads, we knew it would need to be able to process tens of terabytes of data and be deployed anywhere. »
Weka’s core offering is a parallel file system, a type of distributed file system that can distribute and orchestrate data tasks (e.g. copying files) across multiple locations (e.g. servers and workstations) at a time. In addition to this, Weka sells services and capabilities to support AI and machine learning, visual effects and high-performance computing workloads in environments spanning on-premises data centers, clouds public and hybrid clouds.
Zvibel says one of the key benefits of Weka’s architecture is that it can accelerate the training of AI models by reducing the time it takes to copy data between storage locations. “A typical generative AI data pipeline includes multiple stages of copying datasets, wasting vital training time,” he said. “Weka constantly feeds data into the model training hardware so that models can be trained faster. »
Weka competes with data platforms such as DataDirect, Pure Storage, NetApp and Vast Data. Vast is among the most formidable of the group, having closed a $118 million Series E funding round in December 2023, which tripled the startup’s valuation to $9.1 billion.
But Weka appears to be holding its own, with a customer base of more than 300 brands, including AI startup Stability AI, 11 on the Fortune 50, and several undisclosed domestic and foreign government agencies.
Even with a relatively large headcount (around 400 employees worldwide, with plans to increase that number by 25% over the next year), Zvibel said the Silicon Valley-based company now has a “line of focus” for positive cash flow by December 2024. .
“The latest increase was calculated based on favorable market conditions and proactive investor interest, which allowed us to increase on extremely favorable and advantageous terms for Weka,” he added. “Our average burn rate would have to be less than half a million per month before we reach this milestone. We have exceeded $100 million in annual recurring revenue and maintain a hypergrowth trajectory.
techcrunch
