Whereas an rising variety of firms are investing in AI, many are struggling to get AI-powered tasks into manufacturing — a lot much less delivering significant ROI.
The challenges are many. However a generally recurring one is information administration. The info that firms want to coach, run and fine-tune AI fashions is disorganized, siloed, and in any other case unoptimized. In a 2022 survey by Nice Expectations, an open supply information benchmarking platform, 77% of organizations mentioned they have been involved about their information high quality.
Startups that promise to repair these information issues are raking in funding.
On Wednesday, Weka, a platform for constructing information pipelines that deal with a spread of information sources, varieties, and sizes, introduced that it raised $140 million in a two-part ($100 million and $40 million) Collection E spherical led by Valor Fairness Companions, with participation from Nvidia, Norwest Enterprise Companions, Micron Ventures, Qualcomm Ventures, Hitachi Ventures and others. The oversubscribed spherical values Weka at $1.6 billion post-money, double the corporate’s earlier valuation.
Weka CEO Liran Zvibel and Weka’s different co-founders, Maor Ben-Dayan and Omri Palmon, met whereas constructing the information storage startup XIV, which IBM acquired for $350 million in 2007. The trio stayed on with IBM for quite a few years however finally left to pursue different, unbiased ventures.
The issue of managing information continued to gnaw at Zvibel, although, he says.
“I used to be annoyed and disillusioned with seeing clients pressured to make use of disparate, siloed information infrastructure options that have been wasteful, expensive and sophisticated to deploy, handle and preserve,” he mentioned. “The issue grew to become particularly obvious with the rise of cloud computing and the arrival of high-performance computing, machine studying and the earliest AI workloads.”
So in 2013, Zvibel recruited Ben-Dayan and Palmon to construct a brand new set of information instruments — a set which may result in a greater strategy to storing, managing and transferring information.
“We envisioned a platform highly effective sufficient to help the efficiency calls for of next-generation compute {hardware} and large-scale, data-intensive workloads within the demanding and distributed environments,” Zvibel. “To satisfy the wants of contemporary workloads, we knew it might want to have the ability to course of tens of terabytes of information and be deployed anyplace.”
Weka’s core providing is a parallel file system, a kind of distributed file system that may unfold and orchestrate information duties (e.g. copying information) throughout a number of locations (e.g. servers and workstations) without delay. On prime of this, Weka sells providers and capabilities to help AI and machine studying, visible results and high-performance compute workloads in environments spanning on-premises information facilities, public clouds and hybrid clouds.
Zvibel claims that one key benefit of Weka’s structure is that it may velocity up AI mannequin coaching by decreasing the period of time it takes to repeat information throughout storage places. “A typical generative AI information pipeline consists of a number of steps of copying information units, which wastes very important coaching time,” he mentioned. “Weka retains mannequin coaching {hardware} consistently fed with information so fashions will be skilled faster.”
Weka competes with information platforms like DataDirect, Pure Storage, NetApp and Huge Knowledge. Huge is among the many extra formidable of the bunch, having closed a $118 million Collection E funding spherical in December 2023 that tripled the startup’s valuation to $9.1 billion.
However Weka seems to be holding its personal, with a buyer base of over 300 manufacturers together with AI startup Stability AI, 11 of the Fortune 50, and several other undisclosed home and overseas authorities companies.
Even with its comparatively giant headcount (~400 staff worldwide, with plans to develop that quantity 25% within the subsequent yr), Zvibel mentioned that the Silicon Valley-based firm now has a “line of sight” to money move positivity by December 2024.
“The most recent increase was calculated based mostly on favorable market situations and proactive investor curiosity, which enabled us to lift at extraordinarily favorable and advantageous phrases for Weka,” he added. “Our common burn fee is predicted to be lower than half 1,000,000 monthly earlier than reaching that milestone. We’ve exceeded $100 million in annual recurring income and are sustaining a hyper-growth trajectory.”
