Advancing Knowledge Retrieval with Large Language Models
Linde is a leading global industrial gases and engineering company, with a vast global workforce and a large collection of internal technical, policy, and operational documentation in multiple languages.
.avif)
.svg){kind=icon}
Instant Knowledge Retrieval
Natural-language search delivers accurate, source-cited answers in seconds, replacing slow, manual document digging.
Global, Multilingual Access
Employees across regions can query in their own language and retrieve insights from documents written in others.
Enterprise-Grade AI Search
A custom RAG platform hosted on Azure securely processes PDFs, scans, and tables, with automated updates and performance monitoring built in.
The Challenge – Streamlining Knowledge Access
At Linde, teams around the world relied on a fragmented set of internal documents: PDFs, scanned files, multilingual content, tables, and more. As a result, retrieving accurate, relevant information – about equipment specifications, safety protocols, maintenance procedures, internal policies, etc. – was slow, error-prone, and required sifting through multiple systems.
Linde needed a simple, secure, and high-quality solution that would allow employees to “chat” with the company’s knowledge base and get reliable answers – quickly, in natural language, across multiple languages, wherever they are.
That’s when Linde reached out to STX Next to build such a system together.
.avif)
STX Next x Linde – What We Built Together
We approached this not as a mere “search tool,” but as a custom knowledge-retrieval platform built for Linde’s global, multilingual, high-volume document environment. After a thorough analysis of the client’s needs, our dedicated team proposed the implementation of new AI-based tools and features, including:
.avif)
Retrieval-Augmented Generation (RAG) Tool
We implemented a RAG-based solution: when a user asks a question, the system first searches Linde’s internal document database for relevant content; once relevant documents/passages are found, the LLM generates a coherent, context-aware answer to the user. This ensures relevance, context, and reliability.
Data Security & Sovereignty
The entire solution is hosted on Microsoft Azure Cloud, ensuring that all data remains proprietary, stored within Linde’s secure infrastructure – with no external transfers or exposure. Data integrity and security are always our top priorities.
Multilingual & Multi-national Support
Our team designed the system to support multiple languages – reflecting Linde’s global footprint. Employees can query in one language and retrieve answers from documents written in another, breaking language barriers and greatly expanding access.
Advanced Document Indexing & Content Handling
Applying advanced text-splitting techniques.
Adding search-friendly metadata.
Handling tables and structured content separately.
Building automated pipelines to synchronize and update the document database whenever documents are added or modified.
This way, the knowledge base remains up to date and performant without manual overhead.
Source Citation & Reliability Monitoring
To guarantee trust and transparency, the system doesn’t just produce answers – it also cites exactly which documents and passages the answer is derived from. On top of that, we built a framework to monitor chatbot performance continuously – ensuring reliability, compliance, and ongoing quality.
Tech Stack
Business Impact – Increased Speed and Accessibility
By delivering this solution, we helped Linde transform how their global workforce accesses internal knowledge. Here’s the concrete value they gained:
Significantly faster and more accurate search
Employees no longer had to manually dig through multiple search engines, shared drives, or ask colleagues. Instead, immediate, accurate answers surfaced in seconds.
Broader and inclusive information access
Thanks to multilingual support and cross-language document retrieval, staff in any region could access information, regardless of the original document’s language. That broke down information silos and democratized access to knowledge globally.
Better Understanding & Informed Decision-Making
The natural-language interface allowed users not only to get a quick answer but also to ask follow-up questions or request summaries. That improved comprehension and helped teams make better, faster decisions.
Higher Trust & Data Governance
By providing source citations and keeping all data within Linde’s secure cloud infrastructure, the solution reinforced user trust. It ensured data sovereignty – essential for a global engineering company handling sensitive technical content.
Sustainable, Scalable Knowledge Infrastructure
The optimized indexing, metadata tagging, table-handling, and automated update pipelines made the system maintainable over time – meaning knowledge stays fresh and accessible as documentation grows.
Insights and Conclusions
Complex enterprise knowledge = unique challenge
Many off-the-shelf search tools fail when the data is multilingual, heterogeneous (scans, PDFs, tables), or massive. Our approach shows that a custom RAG-based solution can overcome this.
Security and data sovereignty are non-negotiable
For industry leaders like Linde, hosting in a secure cloud with no external data leakage is critical – so enterprise clients need solutions that respect that.
Natural language + source citation + continuous maintenance = trust + usability
It’s not enough to deliver an answer: users need to understand, verify, and trust it. The mix of LLM answer, source, and monitoring delivers on that.
Scalability and sustainability are key
Enterprises evolve; documents get added, language coverage expands. Building pipelines for indexing, metadata tagging, updates – that’s what turns a one-time project into a long-lived asset.
How STX Next Handled The Project
Close partnership
We worked closely with Linde throughout the project – from needs analysis, architecture design, data preparation, to deployment and hand-off.
Cross-disciplinary skills
Our expertise in AI/ML engineering, data engineering, cloud infrastructure, and knowledge of enterprise-grade compliance/security let Linde to achieve their business goals.
Custom approach
The solution we’ve built is tailored to Linde’s complexity: multilingual documents, structured and unstructured data (tables, PDFs, scans), enterprise-grade security, and continuous maintenance pipelines.
.avif)
Are you ready to also transform your business?
Schedule a call with our experts to discover the benefits of partnering with STX Next.
