Al edge firm Blaize sees Australian opportunities
Australian mining and agriculture are two key sectors where NASDAQ-listed tech firm Blaize sees an opportunity for its AI edge inference solutions, a senior global executive has told CommsDay.
Earlier this year the company revealed an alliance with Nokia in the form of a memorandum of understanding targeting Asia Pacific markets for what Blaize describes as practical AI and physical AI systems.
Co-founder and CEO Dinakar Munagala said Blaize’s technology could support Nokia providing value-added services to a range of customers ranging from factories to mines, cities and government that are seeking to invest in Al.
“The commonality is – inferencing is the key,” he told Dinakar Munagala CommsDay. Because Blazie is focused on edge inferencing it supports use cases where customers don’t want their data to leave their jurisdiction or be sent to a thirdparty cloud.
He said he couldn’t yet reveal details of the “next tactical steps with Nokia” but selecting an initial cohort of customers will be key. Example use cases could be facial recognition for immigration services or mine safety, he said.
Senior vice president of marketing Joseph Sulistyo said there were three strategic phases as part of implementing the alliance, beginning with building a joint, enterprise-and carrier-grade reference solution with Nokia.
“We provide production quality inference infrastructure and then Nokia provide essentially their networking stack along with their ability to be able to build their networking analytics,” Sulistyo said. The initial work will be in the form of an AI innovation hub in APAC. The second stage is focused on use cases across different verticals, with Sulistyo flagging the challenges for mining and agricultural production in Australia as potential candidates.
“This kind of two use cases alone basically can build our first use case and validation for a deployable AI inference type of reference or solution stack,” he said.
The third piece is about making the technology available across multiple partner ecosystems.
Munagala said the team that originally formed Blaize were primarily architects from firms such as Intel, Nvidia and Apple. “We built the graphics systems that went into Intel’s laptops, the MacBooks and so on,” he said. “So our DNA is building highly efficient processors and we know GPUs, what they’re good at and how we can complement them for Al.”
Blaize’s platform encompasses both software and hardware, including what it calls the Blaize Graph Streaming Processor, with a focus on inferencing. AI development has focused on GPUs, Munagala said. “But when you look at- hey, how do you actually operationalise AI? You do need certain cost, power metrics.” Blaze is focused on “physical AI infrastructure.” “And for that you can’t pack a gigantic chip or a server into every physical use case,” he said.
Rohan Pearce