Blog

: Modern buildings are no longer just concrete, glass, and steel. They are living systems—constantly reacting to weather, occupancy, and human behavior. At the heart of these systems lies HVAC infrastructure, quietly responsible for comfort, productivity, and a surprisingly large share of global energy consumption. Across the European Union, buildings are one of the biggest energy and climate levers....

After we explained in the first part of our series how LLM-based OCR fundamentally differs from classic methods, and in the second part examined the technical implementation, we now turn to the crucial question of model selection.. The market for open-source models is moving rapidly, and the choice of the right "engine" significantly determines the quality and efficiency of the pipeline.

In this second part, we focus on the practical implementation of this high-performance pipeline. We show step-by-step how to set up a dedicated, fast processing server on your own NVIDIA GPU using Podman (on Rocky Linux) and the vLLM inference engine. We then build an asynchronous Python client to fully leverage the GPU's power and process even large stacks of documents.

In this first part we look at the conceptual advantages of Large Language Models (LLMs) in document analysis. The technical implementation and practical code examples of the two contrasting pipelines will follow in detail in an accompanying article.

It wasn't long ago that computers could only repeat speech, forming sentences independently was completely out of the question. It's obvious to everyone that this is no longer the case. The days when computers only spoke like parrots are numbered. When you type a prompt into the input field of one of the countless chatbots, the question of how the input is understood, or whether it is truly understood in the real sense, becomes almost inevitable. How does the model understand the term “tree”, for example, and how does this understanding differ from human understanding?