Stölzle Lausitz: Turning Shop Floor Experience into a Digital Asset
Revolutionizing Glass Production: How Stölzle Lausitz Codified Tribal Knowledge into a Real-Time AI Assistant
About the Customer
Stölzle Lausitz produces high-end crystal glassware by combining traditional craftsmanship with industrial automation. In high-precision manufacturing, even small technical faults can disrupt production cycles. To stay competitive, the company focuses on operational excellence and finding ways to preserve decades of glassmaking expertise.
As the lead architectural partner, happtiq designed a system that connects industrial shop-floor data with Google Cloud’s AI capabilities. We built the Production Line Assistant to turn fragmented logs into a verified knowledge base.
Our team handled the full implementation, from setting up high-performance local data pipelines to integrating Gemini Enterprise Agent Platform for real-time recommendation logic. By combining on-site reliability with Gemini's intelligence, we helped Stölzle Lausitz transform manual troubleshooting into a structured digital asset.
Partner Role in the Project
Stölzle Lausitz faced a data continuity problem. While production lines generated clear error codes, the solutions were trapped in fragmented spreadsheets and the unwritten expertise of senior staff. An initial attempt to use LLMs to organize this data lacked the precision needed for a factory setting.
This created several bottlenecks:
Operators frequently had to figure out solutions from scratch instead of using proven fixes.
Valuable work-arounds discovered by operators on the fly were never recorded, meaning this collective intelligence vanished at the end of every shift.
Downtime durations varied because there was no standard troubleshooting manual.
Reliance on manual exports and Excel lists made it impossible to provide real-time guidance during a live production halt.
The Challenge
Our goal was to bridge the gap between old-school industrial processes and modern cloud intelligence.
The system uses a containerized backend with Python FastAPI and a PostgreSQL database for fast data retrieval. We integrated Google Cloud’s Gemini Enterprise Agent Platform and the Gemini 2.5 Flash model to act as a validator. This AI layer reads raw feedback from operators and turns it into structured, verified steps for the company’s "institutional memory".
To keep the system accurate, we built an Admin Control Center. This allows senior specialists to refine AI suggestions and set priorities when historical data is limited. The engine uses a custom ranking formula that weighs feedback, manual boosts, and success ratios to give operators the best advice first.
The Solution
The Production Line Assistant changed the production floor from a reactive environment to one driven by data.
Faster Fixes: Technical faults are matched against a dynamic catalog of solutions, providing operators with ranked recommendations to cut repair time.
Capturing Shop Floor Wisdom: Gemini automatically structures informal operator notes into formal documentation, preserving expertise without increasing the administrative load.
Data Integrity & History: By moving from spreadsheets to a high-performance database with SCD Type 2 versioning, we ensured a "single source of truth" that tracks how solutions evolve over time.
Scalable Architecture: The production-ready stack is designed for rapid duplication, allowing new manufacturing lines to be onboarded in hours.
The Result
Partner Bio
At happtiq, the cloud consulting & integration company, we adopt, implement and leverage the cloud following best-practices for our customers.
Find out more in our One Pagers
