California-based IoT development platform Golioth has announced an AI-ready IoT infrastructure for microcontrollers, designed to streamline AI training, deployment and model execution on IoT data, both on devices and in the cloud.

With this release and new partnerships, Golioth is empowering enterprises to integrate AI efficiently into their operations through AIoT, to innovate and improve products faster.

The IoT industry is notably complex, with over four trillion decision permutations facing embedded engineers. As AI adoption increases across industries, Golioth’s AI capabilities aim to simplify integrating AI into IoT products by reducing the decision-load with flexible AI infrastructure.

Features include streaming large unstructured data files such as images and audio, targeted over-the-air (OTA) updates for AI models, and integrations with AI platforms.

“When IoT developers are still struggling with basic IoT infrastructure problems like connectivity and device security, AI – while a very real business priority – can really slow down IoT projects,” said Golioth CEO Jonathan Beri. “Golioth simplifies IoT development by providing IoT infrastructure as middleware. Adding AI into the development workflow in a low-friction way positions Golioth as the platform of choice for building AI-enabled applications, making it easier for companies to deploy, manage and continuously improve AI models with IoT data.”

Golioth supports streaming rich media data such as images and audio from devices to the cloud for analysis and training. This capability allows developers to use high-resolution images from AI-connected cameras in computer vision applications, or audio-clips from security sensors or natural language processors, for analysis.

Partnering with Edge Impulse, Golioth can stream IoT data directly for model training, unlocking data sources for creating and refining AI models.

“The integration of Edge Impulse and Golioth takes the complexity out of deploying edge AI models and IoT tools,” said Zach Shelby, CEO of Edge Impulse (edgeimpulse.com). “It’s a great match for solving real use cases, from predictive maintenance to environmental monitoring to cargo tracking and more, and will drive exciting innovation in many areas.”

Golioth also supports object storage destinations including AWS S3, Google Cloud Storage and Azure Blob Storage, providing flexible and scalable options for storing and analysing large datasets.

Golioth’s OTA system now supports larger artefacts and a broader range of asset types, including AI models and media files. This enhancement simplifies the deployment of AI models to devices, ensuring models remain up-to-date with the latest capabilities, facilitating efficient model deployment and updates.

“Golioth’s flexible OTA updates and large image upload capabilities have accelerated our product development,” said Tomas Beklin, CTO of Flox Robotics (floxrobotics.com). “These features allow us to quickly deploy AI deterrent systems to keep wild animals away from areas where they are not supposed to be.”

Golioth Pipelines now supports running AI models on live IoT data, integrating with AI platforms such as Hugging Face and Replicate, as well as OpenAI and Anthropic. This allows companies to perform AI analysis on IoT data both on devices and in the cloud, enabling AI-driven insights and automation by routing results to any cloud destination. This integration also lets companies leverage AI in the cloud without modifying existing devices or requiring special hardware to run AI models at the edge.

With the recent explosion of generative AI, Golioth has been evaluating the space for practical options.

“We wanted our AI tools to truly empower our customers,” said Golioth head of product Dylan Swartz. “By interviewing them, we identified their real needs and built a suite of capabilities tailored to those needs, ensuring practical and impactful AI that benefits sustainability initiatives and our planet.”

Clive Pinfold, CEO of Heatr, a company focused on optimising New Zealand’s energy grid, added: “With Golioth, we are building AI-enabled devices for energy grids that optimise energy use and contribute to a greener planet today. The platform’s robust Pipelines feature allows us to stream live data to our cloud AI models, enabling rapid innovation and more efficient, sustainable energy consumption.”

For companies with existing expertise in AI and building AI-enabled products, Golioth provides tools to streamline processes. For organisations new to AI, Golioth simplifies AI integration without the need to build AI infrastructure, allowing them to leverage AI functionalities with little effort.

Innovation in the hardware sector is also supporting the growth of AI.

“There’s going to be an explosion of low-power AI capable microcontrollers soon embedded in everything,” said Kwabena Agyeman, CEO of OpenMV (openmv.io), a company focused on making high-performance, low-power microcontroller boards. “New chips are hitting the market that are 100 to 300 times faster in real use cases than microcontroller systems before. We’re happy to see Golioth building out the pipelines necessary for these new systems to reach their full potential.”

In support of the launch, Golioth is offering a $1000 credit to new customers until October 31, 2024. As a bonus, new customers that launch any size fleet and start streaming within 30 days of activation can enjoy an additional $1000 credit on any AI-related data destination. In addition, Replicate is offering a $500 credit to any new customers from Golioth.

Founded in 2020 by CEO Jonathan Beri, Golioth (www.golioth.io) helps engineers overcome the problems of IoT development to get to production, fast. The platform, firmware SDK, suite of tools and reference designs can accelerate the time to market for devices making IoT infrastructure more secure, dependable and sustainable.