Laneyes Launched Argos to Accelerate Edge AI Deployment

To achieve the vision of distributed intelligence and real-time automation, enterprises are seeking to deploy Artificial Intelligence (AI) as close to the location of data processing workload as possible. Ideally, this means AI in all edge devices, ranging from large connected vehicles to tiny sensors in electrical and mechanical equipment. However, the realization of such a vision is met with multiple challenges. The diversity of edge devices means any enterprises that want to use custom AI solutions in their devices needs to identify, develop, and trial the right interface, middleware, framework, and cloud solutions that are optimal for edge AI. This requires investing in an AI development team with skillsets in data science, firmware and software technologies, app interface, and full stack development, resulting in heavy investments and long development cycles. This is also the reason why many machine learning projects are struggling to move past the testing and trialing phase, and those that make it through struggle with operationalization, diagnostics, monitoring, and maintenance.

In March 2020, San Diego, California-based startup Laneyes launched Argos, its Machine Learning Operation (MLOps) solution for vision-based AI deployments. One of the key advantages of Argos is the speed of operationalization. Traditionally, edge AI deployment takes 12 to 16 weeks. According to Laneyes, Argos can shorten that deployment time to less than a week, accelerating the time to market for AI. Such speed is crucial for enterprises that want to minimize their opportunity cost.

In order to quickly provision and deploy edge AI, Argos features ready-made machine learning templates for most machine vision use cases. Enterprises that have an internal AI development team can bring their own devices and models. Argos is able to generate inference engine for models based on all major AI frameworks, namely TensorFlow, PyTorch, and ONNX, and the engine runs on Linux-based devices that are powered by NVIDIA’s Graphics Processing Unit (GPU), Google’s Coral, and Intel’s Central Processing Unit (CPU), Field Programmable Gate Array (FPGA) and Vision Processing Unit (VPU). During the deployment phase, Argos’ provisioning tools support zero code web user interface-based deployment through a secure web portal hosted in either a public cloud or a private cloud for users that have stringent data security and privacy requirements. Once deployed, enterprises have the advantage of cloud-based device monitoring, orchestration and management, alert management, and machine learning model performance monitoring and retraining. All the information exchanges are based on the REST interface on the cloud for sending or receiving messages from each device in the field.

By offering a one-stop solution for AI deployment, Laneyes’ Argos enables developers to develop, deploy, and maintain vision-based AI models. Laneyes is currently targeting verticals with established brownfield infrastructure in image and video capturing, such as the manufacturing and surveillance market. Legacy inspection machinery in a factory with a camera for vision inspection can be retooled to add AI capabilities that can be controlled from the cloud. In addition to vision inspection, the startup’s strength in vision-based AI allows it to explore opportunities in other verticals, such as retail, transport & logistics, and smart cities, where camera and sensor adoption have been growing rapidly for parcel counting, planogram, or automatic license plate recognition-based gate management systems.

Needless to say, many companies are targeting the low-code or zero-code AI deployment market. Some developers may prefer development tools and infrastructure from public cloud providers like AWS SageMaker, Google Cloud AI, and Azure AI to build and deploy edge AI, while having control over the full deployment pipelines. Other companies provide low-code edge AI deployments (like Balena) or provide custom tools to compress machine learning models for edge AI devices (such as Edge Impulse, Nota, and SensiML). In addition, there are other vertically-focused AI-as-a-Service startups. Falkonry and One Tech focus heavily on the factory environment, while Neurala and Nokia Software IoT target other asset-heavy verticals, such as robotics automation, oil & gas, utilities, and maritime.

This is still early days for edge AI, with no clear market leader or winning formula. ABI Research estimates the total installed base of edge AI device to reach 5.2 billion globally by 2025, with a Compound Annual Growth Rate (CAGR) of 18% from 2020 to 2025. Heterogeneity in the edge AI landscape means Laneyes’ one-stop solution will resonate with both small and medium enterprises and large companies. Laneyes can draw on its strength from its hardware- and software-agnostic approach in vision-based AI, being able to support all major AI frameworks and major edge AI chipsets. Moving forward, the market may consolidate around several key players with incumbency advantages or strong domain expertise. ABI Research believes that the right partnerships and collaboration with system integrators or equipment vendors will unlock future successes for Laneyes.