Welcome to DatMap Platform

2026-06-05 · @datmap.ai

Research moves fast. Thousands of papers and patents appear every month, and reading lists don't scale. Data maps offer a different way in: embed every document, project the embedding space to two dimensions, cluster it hierarchically, and let an LLM name what each cluster is about. The result is a zoomable landscape of an entire field.

What you can do here

The pipeline, briefly

  1. Documents → sentence-transformer embeddings
  2. Embeddings → 2-D layout (UMAP)
  3. Layout → hierarchical clusters (HDBSCAN-family)
  4. Clusters → human-readable names (LLM)
  5. Everything → an interactive HTML map you can pan, zoom and query

More deep dives coming soon.