The convergence of artificial intelligence and the Internet of Things (AIoT) represents one of the most groundbreaking technological advancements of today. Imagine an ecosystem where devices communicate, adapt, learn, and improve over time. This network capable of self-optimization and intelligent decision-making is exactly what Golioth, an IoT platform built for modern development teams working on the tiniest of devices, envisions and brings to life.
Golioth is a comprehensive suite of cloud services for embedded devices, acting as a universal connector that bridges hardware and cloud functionality. It’s known for helping developers focus on innovation instead of concerning themselves with connectivity and infrastructure. How? By facilitating secure connectivity, enabling efficient over-the-air (OTA) updates, and offering robust device and data management.
It’s this specialization that allows Golioth to acknowledge that the fusion of AI and IoT holds immense promise in reshaping industries and, at the same time, introduces several challenges. IoT development alone is already complex. Embedded engineers face over four trillion decision permutations when designing smart devices. Imagine adding AI into the mix. Training and deploying AI models require massive datasets, robust version control, and efficient inference mechanisms. These are factors that are difficult to manage in distributed, real-world deployments.
Golioth, recognizing these hurdles, launched AI-ready IoT infrastructure for developers. “IoT developers are already struggling with issues in infrastructure like connectivity and device security. Leveraging AI should enhance their projects and not slow them down. Golioth provides an IoT infrastructure that simplifies this process. It acts as a middleware solution that streamlines AI integration. This is how we help companies easily deploy, manage, and refine their AI models in real-world IoT environments,” Jonathan Beri, Golioth founder and CEO, explains.
The platform’s strategy focuses on three core pillars, starting with training data. IoT devices generate vast amounts of raw information every day, yet much of this data remains underutilized due to issues of accessibility and improper formatting. Golioth enables developers to transform device data into cloud native data formats that can now be stored, analyzed, and integrated with major object storage services like AWS S3, Google Cloud Storage, and Azure Blob Storage. Data sets can then be used to train the next generation of models, which in turn, can be brought back to the device, completing a cycle of continuous improvement and innovation that is the promise of AI.
Golioth has also established a partnership with Edge Impulse, a leader in edge AI development. This collaboration simplifies streaming data from Golioth’s platform directly to Edge Impulse, where it’s prepared for advanced model training specifically optimized for microcontroller-class devices. “Our partnership with Golioth opens up practical solutions for real-world challenges, from predictive maintenance and environmental monitoring to cargo tracking. We hope to catalyze innovation across diverse industries,” Zach Shelby, CEO and co-founder of Edge Impulse, remarks.
After training data, the next pillar of Golioth’s approach is model management through flexible OTA updates. Deploying AI models to multiple devices without incurring significant downtime or wasting resources is critical. Golioth’s enhanced OTA update system supports various artifact types (including AI models and media files), allowing for rapid, targeted updates without needing a complete firmware overhaul. This innovation conserves bandwidth and battery life and ensures that devices continuously operate with the most current AI models. The Flox case study perfectly demonstrates this capability.
Flox’s mission is to protect animals from harm across remote and varied terrains. The company needed a solution that could manage vast volumes of unstructured data (from sensor readings, high-resolution images, etc.) and reliably update AI models across devices. Leveraging Golioth’s OTA update system, Flox was able to remotely deliver AI model updates without disrupting device operations, even in low connectivity areas.
The solution incorporated Zephyr-based applications on nRF9160 boards, communicating securely with Golioth’s cloud infrastructure. What did this robust setup enable? Real-time data steaming back to the cloud for ongoing model refinement while ensuring that the latest AI models are deployed swiftly and securely. Overall, it has enhanced the reliability and efficiency of Flox’s animal deterrence systems.
Following training data and model management is the third pillar: inference, the engine that drives real-time AI decision-making. Golioth’s platform supports both on-device inference (essential for applications demanding immediate responses) and cloud-based inference for tasks that require substantial computational power and detailed analysis. This dual capability ensures that Golioth provides the infrastructure needed for decisions that must be made instantly at the edge and beyond or processed through more complex algorithms in the cloud.
It’s worth noting that with Golioth Pipelines, AI models can now integrate with platforms like Replicate, Hugging Face, and Anthropic — with more to come. This allows developers to perform inherence directly within their data pipelines.
Golioth for AI is a significant step in seamlessly integrating AI with IoT. It provides developers with a toolkit to turn raw IoT data into intelligent insights by addressing the challenges of data acquisition, model management, and inference. Beri states: “Our goal was to build AI tools that truly empower our customers. We developed a platform that makes AI practical, scalable, and impactful for IoT across industries because we understood their real-world needs.”
