Leo Platzer

Starting from a real document corpus full of overlapping, look-alike files, we walk through what it takes to make retrieval on those files reliable, from deduplicating to enriching with metadata. Watch how each step reshapes the vector space, and what happens to the answers that come back.