Return to New York on June 5th to work with executives to explore comprehensive ways to audit AI models for bias, performance, and ethical compliance across diverse organizations.Find out how to participate here.
Data is the lifeblood of modern AI systems, but making it available to these complex workloads is extremely difficult. today, wekais a startup working to simplify this challenge by making data continuously available on-demand with its own AI-native services, and has raised a Series E round of 1 on the back of strong customer interest. announced that it had raised $40 million. The investment, all from existing investors, doubled the company's valuation to $1.6 billion. What will happen in November 2022?.
“Few could have predicted how quickly the AI market would rise…But when[generative AI]explodes into the field in December 2022, there will be global demand for Weka’s data platform software. Demand was surging, with large enterprise customers and research organizations looking to make AI faster,“ Weka President Jonathan Martin told VentureBeat.
The company plans to use the new funding across multiple areas, particularly to further strengthen its platform. It is a software-based solution that eliminates data bottlenecks caused by legacy architectures and creates „dynamic pipelines“ that continuously feed data to GPU and AI workloads. This increases efficiency and sustainability.
What does Weka bring to the table?
Despite company leaders' repeated commitment to modern workloads such as generative AI, downstream teams struggle to deliver these projects due to gaps caused by data silos and legacy architectures. I'm having a hard time with it.Typical Generative AI pipeline It revolves around multiple steps of copying datasets, which creates a bottleneck that slows down the training process and consumes more energy.
Founded in 2013, Weka solves this problem with what it calls „dynamic data pipelines.“ Essentially, the company's software-based data platform leverages a unique zero-copy architecture that eliminates time-consuming copies and speeds up each step of the AI pipeline to ensure GPUs are constantly fed with data. Masu. This ultimately allows you to train models faster and more efficiently, reducing time to insight and improving business outcomes.
“By enabling organizations to simplify their IT stacks to support demanding AI and GPU-intensive pipelines, delivering significant cost savings and speed, Weka customers can do it faster than their competitors for less. The significant performance improvements also result in significant savings in the power required to run GPU servers, making Weka the most sustainable way to implement large-scale AI projects. ” Martin explained.
At the heart of the Weka Data platform is a scale-out shared parallel file system called WekaFS. Connects directly to Peripheral Component Interconnect Express (PCIe) attached Non-Volatile Memory Express (NVMe) drives. It handles a variety of data types, sizes, and IO profiles, delivering 10x the performance of traditional network attached storage (NAS) systems and 3x the performance of local server storage.
“The Weka Data Platform is designed to address complex data challenges and demanding data environments, including large enterprises, cloud service providers, research institutions, media companies, AI/ML companies and startups, and financial services companies to improve IoT applications and performance. and intensive next-generation workloads such as AI, ML, HPC, quantum computing, 16K media, and VFX,” Martin explained.
On the sustainability front, the platform's ability to use dynamic pipelines to improve GPU utilization will help customers save 260 tons of CO2e per petabyte of data stored. the president insists.
Great growth and the road ahead
on the other hand, The wave of gen AI sparked after the rise of ChatGPT Almost two years ago, Weka has been preparing for this era since its inception. As a result, the company (like other companies) aggressively sells to customers that come in front of them, rather than evolving their products to meet market needs.
“We are already focused on modernizing the enterprise data stack by designing solutions that can support the speed, scale, simplicity, and sustainability requirements of modern performance-focused workloads such as AI/ML. We were not only prepared for this change, we were ahead of the curve,” Martin said.
The company currently has more than 300 customers, including 12 of the Fortune 50 companies. Prominent AI companies using Weka's platform include Stability AI, Midjourney, Eleven Labs, The Center for AI Safety, and AI service providers/GPU clouds like Iris Energy (IREN). ), Applied Digital, NexGen Cloud, Yotta. On the financial front, annual recurring revenue from the company's software subscription model has doubled year over year and now exceeds $100 million. Martin predicts that number will triple or even quadruple this year.
With this funding, Weka increases its cash reserves from previous rounds and seeks to scale its business to meet the demand for AI infrastructure due to the generative AI boom. Martin said this includes investing in research and development, strengthening its data platform and investing in customer success initiatives.
The company expects to grow its workforce by at least 25% to 400 people worldwide by the end of this fiscal year. Other notable players that compete with Weka in the distributed file system space are VAST Data, Nutanix, IBM, Dell Technologies, Qumulo, and Pure Storage.