Industrial Sensor Data to Operational Intelligence with Archetype AI on AWS

Industrial environments generate vast amounts of sensor data, but most of it remains underutilized. Deployed on AWS, Archetype AI’s Newton physical world model brings Physical AI to manufacturing, turning raw signals into real-time operational intelligence.

Introduction

The next wave of AI is moving beyond the digital realm and into the physical world. Foundation models have transformed how machines understand language, images, and code. The same paradigm is now being extended to real-world systems through physical world models, AI models that understand the world by learning from sensor signals and observations.

Industrial machines constantly produce signals, vibration, pressure, temperature, electrical current, that encode the state of physical systems. Over the past decade, manufacturers have deployed millions of sensors across factories and industrial assets, generating vast streams of operational data. Yet most of this information remains underutilized. Traditional industrial AI approaches rely on custom machine learning models that struggle to interpret complex, multimodal sensor signals and do not scale across machines and environments.

At the same time, industrial organizations face growing pressure to improve uptime, optimize operations, and address a widening workforce knowledge gap as experienced technicians retire. Improvements in machine performance, energy efficiency, and reliability can have significant economic impact.

A physical world model built for the real world

Archetype AI is building intelligence for the physical world with Physical AI systems that help industrial organizations understand, monitor, and optimize real-world operations using multimodal sensor data. Available through AWS Marketplace, the Archetype Platform combines a physical world model, cloud-to-edge infrastructure, and Physical AI agents, systems that monitor, reason, and act on physical operations in real time, to transform raw machine signals into operational intelligence across manufacturing, energy, construction, and other industrial environments.

At the core of the platform is Newton, Archetype’s physical world model. Newton is designed to understand multimodal sensor data and uncover hidden patterns in any physical system, enabling use cases like predictive maintenance, process optimization, and task verification across operations.

Newton works out of the box, with no task-specific training, to detect anomalies, diagnose conditions, and generate insights, even in environments it has never encountered before. It generalizes across sensors, machines, and conditions, enabling organizations to scale a single model across a wide range of use cases. According to Archetype AI, a customer achieved up to 90% accuracy detecting a range of industrial pump conditions across multiple sites from several labeled examples, no model training was required.

Instead of building and maintaining custom machine learning models for each asset or scenario, organizations can deploy Newton once and scale it across operations. This reduces development time and ongoing maintenance and enables Physical AI to scale across industrial environments.

Figure 1: Newton world model generalizes across sensors, machines, environments, and use cases

Newton is deployed through Newton Agents that can run on the edge, where data is generated and operations take place. Archetype provides prebuilt agents designed for common industrial use cases, including:

• Anomaly Discovery — Detect novel failures and abnormal machine behavior without labeled examples to reduce downtime, lower maintenance costs, and extend asset life.
• Operational State Monitoring — Continuously monitor machine operating states in real time to improve reliability, optimize performance, and identify issues before failures occur.
• Task Verification — Validate worker actions against standard operating procedures to ensure compliance, improve quality, and accelerate onboarding.

Archetype enables manufacturers to operationalize this intelligence across industrial environments, transforming raw machine signals into real-time insights at the edge.

Deployment considerations

The Archetype Platform supports flexible deployment across AWS, including customer-hosted environments. Familiarity with AWS infrastructure, Kubernetes-based deployments, and industrial sensor environments is helpful for understanding the architecture described in this post.

Scaling Physical AI with Archetype and AWS

Archetype AI’s platform and Newton world model can be deployed within customer-managed AWS environments using a scalable, containerized architecture built on Amazon Elastic Kubernetes Service (Amazon EKS). Newton inference services and Newton Agents run as containerized workloads, while managed AWS services support scalable storage, caching, and infrastructure operations. This deployment approach enables manufacturers to operationalize Physical AI workloads within existing AWS environments while supporting high-throughput sensor data processing, real-time inference, and flexible deployment across cloud and edge infrastructure.

Newton understands multimodal sensor signals such as vibration, pressure, temperature, electrical current, radar, and vision. With 8 billion parameters, Newton understands rich representations of physical systems across diverse sensor modalities and provides natural language interpretation. For production deployments, this model can be distilled into smaller inference-optimized models, including an 85-million-parameter model capable of running in resource-constrained edge environments.

Figure 2: Archetype Platform cloud-to-edge architecture on AWS

As shown in Figure 2, the Archetype Platform runs in the customer’s AWS account on Amazon EKS, with Newton Agents deployed in the cloud or in the on-premises and edge environments where operations run. The architecture works as follows:

1. The Archetype Platform deploys from AWS Marketplace into a dedicated Amazon Virtual Private Cloud (Amazon VPC), spanning multiple Availability Zones, in the customer’s AWS account. Container images from Amazon Elastic Container Registry (Amazon ECR) and Newton model artifacts from Amazon Simple Storage Service (Amazon S3) are pulled from the Archetype AI-managed Marketplace account. This is the primary interaction between the Archetype AI-managed account and the customer’s account; all customer data stays within the customer’s account.
2. Amazon EKS orchestrates the platform across CPU and GPU node groups. Newton runs on GPU or CPU nodes based on the workload. GPU nodes also support optional fine-tuning for specialized use cases, though no model training is required. Amazon S3 and Amazon Aurora PostgreSQL-Compatible Edition store sensor data, model outputs, and platform data, and Amazon ElastiCache provides low-latency caching.
3. Prebuilt Newton Agents work out of the box, without task-specific training. Agents can run in the cloud on Amazon EKS, or at the edge, where a distilled, edge-ready Newton model runs in resource-constrained environments. Edge agents can be deployed to on-premises environments and industrial edge hardware, with optional AWS IoT Greengrass or AWS Outposts.
4. Sensor and video data is continuously streamed from PLCs and gateways toedge and on-premises agents directly through existing industrial infrastructure, while data for cloud agents is securely ingested through AWS services such as AWS IoT Core, Amazon Kinesis Data Streams, or Amazon Managed Streaming for Apache Kafka (Amazon MSK).
5. Newton Agents continuously interpret multimodal data to detect anomalies, track operating states, and verify tasks. At the edge, processing happens close to where data is generated. In the cloud, agents analyze data streamed into AWS for centralized, large-scale inference.
6. Newton Agents deliver anomalies, operating states, predictions, and recommendations to existing applications through APIs, supporting maintenance, process improvement, safety, and other industrial workflows.
7. Users access the Archetype web console or APIs to monitor operations and configure agents.

This architecture supports cloud, edge, and hybrid deployments, enabling customers to run Newton Agents wherever their use case requires. On-premises and edge deployments enable local processing for low-latency responses and allow sensitive data to stay on site, while cloud deployments support centralized, large-scale analysis.

Together, Archetype’s Newton model and AWS services provide a flexible cloud-to-edge architecture for Physical AI across industrial environments, enabling use cases from machine health and process optimization to task verification.

From raw signals to real-time insights: Physical AI in action

Using Archetype AI’s Newton world model enables manufacturers and industrial operators to extract insights from complex sensor environments without building custom models for every asset. By combining zero-shot detection (the ability to identify patterns without model training) with multimodal sensor fusion (correlating signals across sensor types such as vibration, pressure, acoustics, electrical current, and radar), Newton can generalize across equipment types and identify hidden patterns in physical systems, even when labeled training data is limited or unavailable.

Optimizing construction site productivity

A global construction leader partnered with Archetype to improve productivity on a large, remote canal expansion project with Physical AI. Limited visibility across dispersed job sites and weather-related disruptions made it difficult to manage operations and prevent delays.

Running on GPU-accelerated Amazon EC2 instances, Newton analyzed 12,000 hours of video alongside historical weather data, applying multimodal fusion to generate daily summaries and identify patterns impacting productivity.

Rapid analysis of equipment utilization, weather, and productivity enabled teams to improve scheduling and resource allocation. Over time, Newton creates a form of institutional memory, helping preserve expertise and enabling more proactive, predictive project management.

Gearbox fault detection in a manufacturing facility

A manufacturing customer needed to detect gearbox failures, such as broken gear teeth, from vibration sensors on motors across the factory floor. The challenge was the diversity of equipment: motors varied widely in size and manufacturer, and only a small number of vibration samples were available for analysis.

Using Archetype’s cloud infrastructure on Amazon EKS, the customer deployed a single world model capable of zero-shot detection across all motor types. Without retraining or fine-tuning, Newton analyzed vibration signals to identify patterns associated with gearbox faults and generalized across different equipment configurations.

Operators accessed the system through a conversational interface, allowing them to ask simple questions such as which motors showed abnormal vibration patterns or early signs of degradation.

The solution achieved 99% fault detection accuracy using minimal data. Newton was deployed 80% faster than traditional approaches while providing a scalable, human-friendly interface for monitoring equipment health across the factory.

Transforming industrial operations with Physical AI

As industrial organizations deploy more sensors and connected equipment, the opportunity to unlock value from machine data continues to grow. However, traditional analytics approaches often struggle to interpret complex, multimodal signals across diverse assets and operating conditions. World models offer a new path forward.

Archetype AI’s Newton world model enables organizations to understand the physical world directly from sensor data, combining multimodal sensor fusion with zero-shot capabilities to detect anomalies, diagnose equipment behavior, and uncover hidden system states without extensive model development or labeled datasets.

Archetype AI and AWS enable organizations to transform raw sensor streams into operational intelligence, bringing a new generation of Physical AI capabilities to manufacturing, energy, and other complex industrial environments.

For information about deploying Newton in your industrial environment, visit Archetype AI or contact their team through the Archetype AI listing in AWS Marketplace to discuss your Physical AI use case.

Watch the 1-Minute Overview Video