Imagine effortlessly extracting key information from dense scientific papers. That's the promise of Collage, a new AI-powered tool designed to revolutionize how we interact with scientific literature. Researchers often struggle to efficiently process the sheer volume of PDFs, especially when dealing with older documents or complex data tables. Collage tackles this challenge head-on by offering a unique, decomposable approach to information extraction. Instead of relying on opaque, end-to-end systems, Collage breaks down the analysis process, allowing users to visualize each step. This transparency not only helps researchers understand how the AI arrives at its conclusions but also empowers them to debug and refine the process. Collage supports a wide array of AI models, from Hugging Face transformers to large language models (LLMs), and provides specialized interfaces for tasks like token classification, text generation, and image processing. This flexibility allows researchers to tailor the analysis to their specific needs, whether extracting chemical compounds, summarizing key findings, or parsing complex tables. What sets Collage apart is its ability to handle the multimodal nature of scientific PDFs. It gracefully integrates text and image analysis, enabling researchers to extract insights from figures, tables, and even older, scanned documents with OCR errors. This is particularly crucial for fields like materials science, where much of the critical data resides in tables and charts. By offering a user-friendly interface and supporting a wide range of models, Collage democratizes access to sophisticated AI-powered analysis tools. It empowers researchers across various disciplines to unlock valuable insights from scientific literature, accelerating the pace of discovery and innovation. While challenges remain, such as potential biases in underlying models, Collage's transparent approach encourages critical evaluation and responsible use. As AI models continue to evolve, tools like Collage will play an increasingly important role in transforming how we interact with and learn from the ever-growing body of scientific knowledge.