It took decades for Machine Learning and Artificial Intelligence to enter the mainstream. Still, it is new enough that it remains the domain of data scientists, machine learning, and software engineers. The challenge is that these experts are in short supply, causing projects to be relatively expensive with relatively long timelines.

Blaize’s AI Studio promises to solve that by allowing subject matter experts, whether that expert is a city planner, doctor, or farmer, to directly create artificial intelligence applications that operate locally. This code-free, visual and open standards tool allows the subject matter expert, with minimal help, to prepare & train, deploy & manage, and use & monitor an edge-based, Artificial Intelligent app.

Although the output is exportable to other platforms, AI Studio complements Blaize’s Graph Stream Processors.  These processors feature low power consumption (e.g. 16 trillion operations per second AI interference with only 7 Watts), allowing processing at the edge, which reduces latency for real-time decision-making. The AI Studio collaboration with various marketplaces allows quick access to models, data, and apps, and knowledge enablement.

AI Deployments – From Months to Days #

To make it easier for the end-users, Blaize has pre-packaged solutions, such as plant disease detection for agriculture, retail security, and human disease detection.

Speaking at Pepcom’s CES2021 virtual event, Dmitry Zaharchenko, Blaize’s VP of Research and Development explains how they applied a pre-trained model to help a physician detect a rare disease. This type of project would normally take months to complete with traditional approaches.

Working with the doctor and processing approximately 12,000 images the AI models achieved 90% accuracy within two days. This greatly exceeds the 76 to 80% accuracy a human would achieve examining x-rays. Zaharchenko believes they could have achieved 99% accuracy, had they spent a couple more days refining the models.

This example illustrates the power of computer vision and AI to leverage the knowledge of a subject matter expert.  For instance, using the above model, a physician’s assistant in a rural area could potentially provide a first-level diagnosis without the patient having to travel to an urban area to see a specialist.

Zaharchenko indicates that this is just the beginning and that they will continue to add use-cases and capabilities, such as Natural Language Processing and audio. Although specific pricing is not yet available, Blaize is looking at three different price levels for AI Studio; Basic, Developer, and On-Premise & Enterprise.

Note: To see an earlier interview with Blaize (when it was known as Thinci) that provides additional insight about their hardware solution that runs the AI algorithms, check out this ViodiTV interview from CES2019.