Mining Novel Data from Large Unlabeled Corpus

An ML system-design prompt for ML engineers: mine new, unique information out of a very large unlabeled corpus, and locate the images that contain specific objects of interest. Only the core task is known; the detailed spec was not pinned down. The Challenge This is an ML system design question for Machine Learning Engineers. The core task: Find novel data: extract new, unique information from a very large set of unlabeled data. Locate objects: find images that contain the specific objects of interest. Notes A domain variant frames the same task on medical data: design a system to mine effective, high-signal training examples out of a large medical corpus. The core skill is identical — define what "novel" / "useful" means, then build the filtering and retrieval pipeline around it.