Turning industrial knowledge into real-time field guidance with GenAI on AWS
Commencis helped Durmazlar centralize image-rich technical knowledge and deliver secure, machine-specific answers to service teams through a serverless RAG assistant built on AWS.
Durmazlar
ABOUT THE CUSTOMER
Durmazlar
Durmazlar is a leading manufacturer of sheet metal processing machinery, producing advanced laser cutting, bending, punching, shearing, and automation solutions for customers around the world. Founded in Bursa in 1956, the company reaches 120 countries through the DURMA brand from 150,000 m² of high-technology production facilities.
As its installed base and service operations continued to grow, Durmazlar wanted technical knowledge to become as accessible and scalable as its machinery. Commencis designed a secure GenAI solution that transforms complex service documentation into real-time guidance within the mobile experience already used by Durmazlar service teams.
THE CHALLENGE
Finding the right answer meant searching across fragmented knowledge
Technical information was distributed across PDF service manuals, Excel error-code lists, shared folders, and the experience of senior personnel. With no centralized knowledge structure, technicians often searched several repositories and document versions before locating the correct procedure.
The challenge was not only finding the document. Important diagrams, machine images, screenshots, and procedural steps had to remain connected to the right technical context. Researching a fault could take approximately 8–10 minutes, extending repair cycles and increasing dependency on central support and experienced colleagues.
PARTNER SOLUTION
A secure, image-rich RAG assistant built for field service
Commencis worked with Durmazlar to build a production-ready assistant that combines multimodal document intelligence, a centralized knowledge base, real-time conversational delivery, and AWS serverless services in one experience.
Technical documentation is transformed into a searchable knowledge layer where diagrams and screenshots stay connected to the instructions they support. Machine-specific retrieval and controlled prompts keep answers grounded in approved content, streamed securely into the service application field teams already use.
Centralized knowledge
Technical manuals, error-code lists, images, and structured metadata are transformed into one searchable knowledge foundation.
Multimodal understanding
Relevant diagrams and screenshots remain connected to the instructions they support, creating richer field guidance.
Grounded RAG answers
Machine-specific retrieval and controlled prompts help the assistant answer from approved technical documentation.
Mobile integration
A secure WebSocket service was integrated into Durmazlar's existing service application with streaming responses and image links.
Cloud Architecture
Serverless, event-driven, and designed for continuous knowledge growth
Amazon Bedrock provides managed access to generative AI and multilingual embeddings. AWS Lambda, Amazon S3, Amazon DynamoDB, Amazon API Gateway, Amazon SageMaker, and Amazon Aurora PostgreSQL Serverless v2 support the document, retrieval, and real-time communication workflows Durmazlar requires.
Secure document operations
Content teams upload, retrieve, update, and delete technical documentation through a controlled interface, keeping the knowledge base accurate and current.
Automated content intelligence
Text, tables, error-code records, diagrams, and screenshots are processed into structured, retrieval-ready knowledge in approximately 4–5 minutes per document.
Shared knowledge layer
Vector search and machine metadata identify the most relevant approved content for each question, keeping answers specific to the right machine.
Real-time RAG experience
Grounded answers and related visuals are streamed securely to the Durmazlar service mobile application, with the first token arriving in about three seconds.
Business Impact
Faster access to knowledge, without sacrificing technical context
The platform reduces the time needed to begin receiving relevant guidance while creating a scalable foundation for future service intelligence.
Outcomes & Benefits
From document search to an intelligent service experience
Accelerate fault research
Technicians can begin receiving relevant, machine-specific guidance in seconds rather than manually searching several repositories for minutes.
Improve consistency
Centralized, approved knowledge helps reduce dependency on individual experience and fragmented document repositories.
Preserve visual context
Relevant diagrams, screenshots, and machine images are delivered alongside procedural guidance when available.
Build a scalable knowledge foundation
Serverless services and managed AI capabilities support continuous document growth without dedicated application-server operations.
