Making machine learning operations easier to use, manage and organize for businesses has always been the goal of the series of best practices known as MLOps.
But while MLOps works well for the necessary processes and core processor-based infrastructure of traditional machine learning, it can also prove useful for more complex deep learning workloads, which can be much larger and more demanding than traditional machine learning requirements.
To fill this gap, New York-based startup Spell has launched what it calls a cloud-agnostic MLOps platform that aims to address the more complex and unique needs of deep learning using the principles. MLOps used for machine learning.
“With deep learning, there aren’t a lot of options, because people are using their own tools,” said Tim Negris, marketing manager for Spell. Corporate AI. But by using the company’s recently unveiled platform, businesses can now more easily manage training, orchestration, monitoring, reporting, their deep learning model dashboard, and more. did he declare.
“It’s basically a data and operations infrastructure,” Negris said of the platform. “He has a database that takes care of everything. For regulatory compliance, in areas like financial services, this is very important. Spell captures data about the models, then catalogs the models and their results, while also tracking information about who created the models and more, he said.
Negris said the Spell platform has received a wide range of improvements over the past two years during its development and is now officially introduced to the market. Spell has been refined using feedback from early adopters to help it mature and focus on helping make deep learning easier for businesses to use.
“Until now, we’ve been in a semi-stealth mode, doing limited promotion and targeted marketing,” Negris said. “But now, over the last six months, we’ve added features and completed the range of features that we want to address. It’s a bit like the coming out party.
While many companies have experimented with AI in their labs or development offices, far fewer companies are currently using AI in production today, Negris said. Spell aims to help improve those numbers by removing barriers to adoption so that it can be brought into real-world business operations, or “operationalized,” with fewer pitfalls.
Spell was co-founded in 2017 by Serkan Piantino – who founded Facebook New York and co-founded Facebook AI Research – and Trey Lawrence, who worked as a leader in technical engineering designing silicon chips, PCBs and firmware, as well as built recommender systems. for e-commerce on eBay and Spring. The company received $ 15 million in Series A funding in January 2019.
“The founders were both involved in building the orchestration and management infrastructure for deep learning at Facebook, eBay and Clarifai,” said Negris. “The aha moment, the thing that came to them and brought them together, was the recognition that these big giant companies can build the infrastructure they need, but for most businesses it’s just too hard. They felt that there was a real opportunity to create an infrastructure management layer under the deep learning workflow.
At its core, Spell automates deep learning workflows, from development to training and deployment to optimization, while strengthening compliance, management and other processes, said Negris.
In addition to its flexibility, Spell can be used by enterprises on-premises or through accounts with multi-cloud and hybrid cloud infrastructure providers, including Amazon Web Services, Google Cloud Platform, and Microsoft Azure, making it independent of the cloud and allows users to choose the best implementation situations for their specific needs, said Negris.
“You have workloads that, due to regulatory and security concerns, are being migrated from the cloud to new on-premises equipment in the data center,” he said. “And conversely, there is a whole class of workloads that are… in many cases experiments, the initial design of the model… [that] is done using on-premise GPUs, which are then migrated to the cloud for the purpose of forming a huge model that can consume many hours of GPU time in many GPUs.
This ability to choose the right place to run models is critical and important to Spell users, he said. “And we’ve also made it possible to seamlessly assemble Spot Instances and On-Demand Instances, which can now potentially cut your costs in half,” he added.
Spell also includes collaboration features to coordinate work between machine learning and data scientist teams, as well as tools for Kubernetes-based autoscaling and enterprise-grade security, secure authentication and controls. user / data access.
Spell customers include Akasha, AlphaSense, Cadmium, Condé Nast, Healx, Mulberry, Originate, Quill, Resemble.AI, Whatnot and Square.
Zohaib Ahmed, CEO of neural-to-speech synthesis provider Resemble.AI, said in a statement that her company is using Spell to orchestrate its production deep learning payloads so that it can focus on building high-performance models. quality. “The flexibility and reliability provided by Spell helps us evolve to create hundreds of models every day,” Ahmed said.
Kevin Krewell, analyst at Tirias Research, said Spell is apparently part of a trend of vendors working to make these tasks easier for business users.
“You’re about to see a wave of companies offering similar MLOps tools,” Krewell said. “For example, Edge Impulse recently released its AutoML EON tool. The ML market is shifting from funding new chip companies to funding new software companies that can bring ML to a wider audience of developers.
And as Spell and other companies jump in to meet those needs, they will seek out their niches, Krewell said. “Because all automatic learning is not the same, companies have the opportunity to specialize. Spell’s expertise lies in deep learning algorithms.
Another analyst, Chirag Dekate, vice president of research at Gartner, agreed that more companies are turning to the market for AI, machine learning and deep learning services.
“Gartner surveys and customer engagements indicate a growing urgency in businesses to operationalize AI,” Dekate said. “Gartner follows hundreds of AI startups, the majority focusing on the various arenas of AI orchestration. Companies are organizing AI platforms to orchestrate, automate, and evolve production-ready AI. “
Spell looks at the situation in a unique way, Dekate said.
“Spell’s approach is differentiated in that it allows companies to leverage a platform approach to operationalize AI,” he said. “[The] The Spell platform enables businesses to improve the productivity of AI teams by exposing underlying resources on-premises or in the cloud as a shared platform for data scientists.
The Spell interface is also “versatile in that businesses take advantage of their familiarity with Python notebooks to get started and use Spell to manage shared resources on-premises or in the cloud,” he said. “The Spell interface also allows easy tracking of projects, metrics and the efficiency of AI pipelines.”