Empowering Shop-Floor Operations with GenAI-Driven Operator Assistance Using AI at the Edge

About Customer

The customer is a technology driven analytics company focused on delivering intelligent, data-powered solutions for the manufacturing and industrial domain. The organization specializes in shop-floor intelligence platforms designed to improve operational efficiency, reduce downtime, and support machine operators with real-time decision making

The manufacturing enterprise sought to enhance shop-floor operations by improving how operators access and use machine-related knowledge during day-to-day production. With a wide range of complex and frequently updated machine manuals, operators needed a faster and more intuitive way to retrieve context-specific operational and troubleshooting information, especially when expert support was not immediately available. To support real-time decision-making in a latency-sensitive shop-floor environment, the platform required AI-driven assistance closer to the machines, enabling reliable responses even in constrained connectivity scenarios through AI deployed at the edge. Additionally, the operating environment called for hands-free, voice-based interactions to support noisy conditions and safety requirements, along with multilingual capabilities to better serve a diverse operator workforce.

Additionally, the Manufacturing Enterprise Platform required a highly resilient and secure system capable of delivering consistent assistance despite intermittent internet connectivity, while maintaining compliance, traceability, and auditability in industrial production environments.

To address operational challenges on the manufacturing shop floor, Comprinno implemented a hybrid cloud edge Generative AI assistant that provides machine operators with real-time, contextual operational and troubleshooting guidance. The solution allows operators to interact using text or voice and receive step-by-step instructions grounded in machine manuals. The architecture combines cloud-based AI services for scalable intelligence with AI at the edge for low-latency and offline support. When internet connectivity is available, queries are processed in the cloud using managed AWS AI services. In scenarios with limited or no connectivity, the same intelligence is delivered locally on shop-floor devices, ensuring uninterrupted assistance and consistent user experience.

By unifying cloud intelligence with edge-based AI, the solution improves operational resilience, reduces machine downtime, and enables faster issue resolution while maintaining security, auditability, and compliance..

Key Actions

Conversational Operator Assistance :

• The AI assistant enables machine operators to ask operational and troubleshooting questions using voice or text in English and regional languages.

• It delivers step-by-step guidance directly from machine manuals, reducing dependency on experts and minimizing downtime.

Knowledge-Driven Troubleshooting (RAG) :

• Machine manuals and troubleshooting documents are digitized, with text, images, and tables embedded and indexed for contextual retrieval.

• Responses are generated using only relevant, validated documentation to ensure accuracy and consistency

Hybrid Cloud and Edge Processing :

• Queries are automatically routed to the cloud or processed locally at the edge based on internet availability.

• This ensures optimal response quality, performance, and cost efficiency across different shop-floor conditions.

AI at the Edge for Offline Support :

• Lightweight AI models run directly on shop-floor devices to handle speech, search, and response generation.

• This enables real-time, low-latency assistance even when internet connectivity is unavailable.

Automated Knowledge Synchronization :

• Knowledge bases and vector indexes are synchronized between cloud and edge environments based on operational needs, ensuring consistent behavior across deployments.

• This ensures operators always access the latest machine manuals and troubleshooting guidance, both online and offline.

Security, Monitoring, and Auditability :

• Role-based access controls, encrypted storage, and secure APIs protect all system interactions and data.

• Centralized logging and monitoring provide end-to-end traceability, support compliance requirements, and enable effective operational oversight.

This architecture represents a secure, scalable hybrid cloud–edge Generative AI solution designed to deliver real-time operator assistance with high availability, low latency, and offline resilience. It combines AWS-managed serverless services with edge-based AI processing to ensure uninterrupted shop-floor support, even in environments with intermittent connectivity.

• Edge Intelligence Layer

AWS IoT Greengrass runs on on-premise edge devices to process operator queries locally when internet connectivity is unavailable.

• Lightweight AI components such as Whisper (ONNX) for speech-to-text, MiniLM for embeddings, FAISS for vector search, and TinyLlama for response generation enable low-latency, offline AI inference.

• When connectivity is restored, the edge environment automatically synchronizes vector databases and operational data with Amazon S3 to maintain consistency across cloud and edge deployments.

• Application & Compute Layer

Amazon API Gateway securely receives user queries and routes them to AWS Lambda functions deployed across multiple Availability Zones for high availability.

• AWS Lambda orchestrates request handling, response processing, and integration with AI services in a fully serverless, event-driven architecture, ensuring automatic scaling and reduced operational overhead.

• Generative AI & Knowledge Retrieval

Amazon Bedrock powers cloud-based large language model inference, generating accurate, context-aware responses using enterprise knowledge bases.

• Retrieved responses are grounded using synchronized vector stores, ensuring that AI outputs are relevant, reliable, and aligned with validated documentation.

• For voice interactions, SarvamAI handles speech-to-text and text-to-speech processing, enabling natural, multilingual conversational experiences.

• Data Storage & Synchronization

Amazon S3 serves as the centralized storage layer for vector database backups, historical interactions, and synchronization artifacts from edge environments.

• This enables seamless knowledge updates and supports reliable disaster recovery and version control of AI knowledge assets.

• Security & Compliance

AWS Identity and Access Management (IAM) enforces fine-grained access control across cloud and edge services.

• AWS Key Management Service (KMS) encrypts sensitive data at rest and in transit, while AWS Certificate Manager secures service communications.

• AWS CloudTrail and AWS Security Hub provide audit logging, threat detection, and compliance visibility across the entire architecture.

• Monitoring & Observability

Amazon CloudWatch captures logs, metrics, and alarms for Lambda functions, API Gateway, and system health.

• Centralized monitoring ensures operational visibility, faster troubleshooting, and proactive issue detection.

• Resilience & Fault Tolerance

The architecture leverages multi-AZ serverless deployments in the cloud and edge-based fallback processing to ensure continuous operation.

• This hybrid design delivers high availability, fault tolerance, and consistent operator support regardless of network conditions.