With businesses debating whether to host and run the latest wave of AI in the public cloud, in a private cloud, or on-premises, Cloudera is betting on a bit of all of the above.
The company, which offers a hybrid approach to data storage, recently collaborated with Nvidia to roll out a new service that makes it easier for enterprises to build and deploy AI models as the 16-year-old platform continues to remake itself for the AI age. The move, announced at a conference in New York this month, comes on the heels of Cloudera’s June acquisition of Verta, a startup that helps customers manage machine learning models.
Cloudera is building up these tools as CEO Charles Sansbury claims a shift is taking place in how large companies think about their computing needs. Energy and compute costs are driving more global corporations to run generative AI applications on-premise rather than in the cloud, he said.
“A year ago, if you’d asked these large global customers, ‘What does your endpoint computing architecture look like in five years?’ they would have said, most of my workloads are moving to the cloud. Just in the past year, that tune has changed dramatically,” Sansbury told Tech Brew. “Generative AI-based applications for large companies will run on on-premises hardware, not on-cloud, for purposes of control, security, but also cost.”
Model management: Rather than building and offering its own large language models (LLMs), Cloudera’s AI strategy focuses on giving enterprise customers more tools to build and manage their own models with their data on the platform.
Cloudera’s new AI Inference service taps Nvidia Inference Microservices (NIM), an offering that packages hardware and software so that developers can more easily build and deploy industry-specific chatbots, augmented search engines, and agents on their own servers.
Keep up with the innovative tech transforming business
Tech Brew keeps business leaders up-to-date on the latest innovations, automation advances, policy shifts, and more, so they can make informed decisions about tech.
Nvidia’s ultimate vision for services like these is to push foundation models beyond chatbots to agents that function almost as employees of a given company, according to Kari Briski, VP of AI software product management at Nvidia.
“Think of a foundation model as new employees. The foundation models are trained on the internet of the world. They are very smart. They’ve graduated. You hire them, but they need to be onboarded to your company, your enterprise,” Briski said on stage at the conference. “They need to learn your vernacular. They need to learn your systems.”
Seventh-inning stretch: For now, however, Sansbury said companies remain mostly in “the sixth or seventh inning of getting their data estates in shape.” Companies are still searching for an all-important “killer app,” he said.
“There’s been a ton of spending in doing the infrastructure piece and the data science and the data ranking piece to get your data ready. But we’re still very early on in corporate adoption of AI,” Sansbury said. “The corporate governance folks are saying, ‘OK, we’re giving you all this money. Where’s our return on investment?’ So, they’re looking for quick-win use cases.”
Some of those “quick wins” include customer support, coding assistance, and content generation in areas like marketing and legal, Sansbury said. But more transformative uses could take more data and infrastructure alignment and overhaul, he said.
“The issue is that right now, it’s a little bit dependent on the quality of the data scientists that you have, and the data scientists are being forced to get closer to understanding the business problems,” Sansbury said. “So, we’re in this early primordial soup where we've now got better control over the data and the analytics, but we need kind of the human side of business understanding to develop.”