California software company Blaize has developed a platform to help developers take artificial intelligence (AI) from idea to deployment and beyond.

The company’s AI Studio is an open and code-free software platform that spans the complete edge AI operational workflow from idea to development, deployment and management. It is said to reduce edge AI application deployment complexity, time and cost.

The platform does this by breaking the barriers within existing application development and machine-learning operations (MLOps) infrastructure that hinder edge AI deployments. Eliminating the complexities of integrating disparate tools and workflows, along with the introduction of multiple ease-of-use and intelligence features, it can reduce from months to days the time required to go from models to deployed production applications.

“While AI applications are migrating to the edge with growth projected to outpace that of the data centre, edge AI deployments today are complicated by a lack of tools for application development and MLOps,” said Dinakar Munagala, CEO of Blaize. “AI Studio was born of the insights to this problem gained in our earliest edge AI hardware customer engagements, as we recognised the need and opportunity for a new class of AI software platform to address the complete end-to-end edge AI operational workflow.”

The combination of innovations in user interface, use of collaborative marketplaces, end-to-end application development and operational management collectively bridge the operational chasm hindering AI edge RoI, claims the company.

“AI Studio is open and highly optimised for the AI development landscape that exists across heterogeneous ecosystems at the edge,” said Dmitry Zakharchenko, VP of research and development at Blaize. “With the AI automation benefits of a truly modern user experience interface, AI Studio serves the unique needs in customers’ edge use cases for ease of application development, deployment and management, as well as broad usability by both developers and domain expert non-developers.”

Deployed with AI edge computing hardware offerings that address unserved edge hardware needs, AI Studio can make AI more practical and economical for edge use cases where unmet application development and MLOps needs delay the pace of production deployment.

“In our work for clients, which may include developing models for quality inspection within manufacturing, identifying stress markers to improve drug trials or even predicting high resolution depth for autonomous vehicles, it is vital that businesses can build unique AI applications that prove their ideas quickly,” said Tim Ensor, director of AI for Cambridge Consultants. “AI Studio offers innovators the means to achieve this confidence in rapid timeframes, which is a really exciting prospect.”

The code-free visual interface is intuitive for a broad range of skill levels beyond just AI data scientists, which is a scarce and costly resource for many organisations.

“Hey Blaize” summons a contextually intelligent assistant with an expert knowledge-driven recommendation system to guide users through the workflow. This ease of use can enable AI edge app development for wider teams from AI developers to system builders to business domain subject matter experts.

Users can deploy models with one click to plug into any workflow across multiple open standards including Onnx, Open VX, containers, Python or GStreamer. Support for these open standards means the platform can deploy to any hardware that supports the standards.

Marketplace support allows users to discover models, data and complete applications from anywhere – public or private – and collaborate continuously to build and deploy AI applications.

It supports open public models, data marketplaces and repositories, and provides connectivity and infrastructure to host private marketplaces. Users can continually scale proven AI edge models and vertical AI services to reuse across enterprises, choosing from hundreds of models with drag-and-drop ease to speed application development.

The model development workflow lets users train and optimise models for specific datasets and use cases, and deploy quickly into multiple formats and packages. With the click of a button, a transfer learning feature retrains imported models for the user’s data and use case.

An edge-aware optimisation tool, NetDeploy, automatically optimises the models to the user’s accuracy and performance needs. Users can build and customise complete application flows other than neural networks, such as image signal processing, tracking or sensor fusion functions.

As an end-to-end platform, it helps users deploy, manage, monitor and continuously improve their edge AI applications. Built on a cloud-native infrastructure based on microservices, containers and Kubernetes, it is said to be scalable and reliable in production.

In smart retail, smart city and Industry 4.0 markets, users can realise new levels of efficiency in AI application development and deployment. Examples include: complete end-to-end AI development cycle reduction from months to days; reduction in training compute by as much as 90%; edge-aware efficient optimisations and compression of models with a less than 3% accuracy drop; and contextual conversational interfaces that eclipse visual UI.

The platform is available now to qualified early adopters, with general availability early next year. The offering includes licences for individual seats, enterprise and on-premise subscriptions, with product features and services suited to the needs of each licence type.

“We are seeing tremendous benefit for early adopters,” said Rajesh Anantharaman, Blaize senior director

Blaize has secured $87m in equity funding to date from Denso, Daimler, Sparx Group, Magna, Samsung Catalyst Fund, Temasek, GGV Capital, Wavemaker and SGInnovate. With headquarters in El Dorado Hills in California, Blaize has teams in Campbell in California and Cary in North Carolina, with subsidiaries in Hyderabad (India), Manila (Philippines), and Leeds and Kings Langley (UK), with more than 300 employees worldwide.