For a more readable view with labeled companies and meaningful axes, see the positioning chart →

Exploratory map (t-SNE) · Conference map →

Company map

Every company present at four or more conferences, placed so that companies showing up at the same conferences sit near each other. Colors are clusters: groups of companies that work the same circuit.

Cluster

Clear

2,034 companies. Dot size: number of conferences present at. Hover a dot for details and to light up its cluster; click to open the company. Showing 138 matches in cluster 0.

Selected cluster

NeurIPS · ICML circuit

138

Characteristic conferencesNeurIPS, ICML, ICLR, CVPR, ACL, MLSys

Top categoriesNot a vendor (72), GenAI & LLMs (16), Cloud infrastructure (13), Data providers (8)

Companies in this cluster

0Meta77 conferences
0Netflix50 conferences
0Uber50 conferences
0Apple48 conferences
0JPMorgan Chase46 conferences
0Adobe43 conferences
0LinkedIn43 conferences
0Stanford University39 conferences
0Capital One36 conferences
0Alibaba Cloud34 conferences
0Huawei28 conferences
0New York University25 conferences
0Intuit24 conferences
0Two Sigma24 conferences
0ByteDance23 conferences
0Lyft23 conferences
0Tencent22 conferences
0Cohere21 conferences
0Citadel19 conferences
0Samsung Electronics19 conferences

Top 20 of 138 by conferences present at.

How to read this

The axes

They mean nothing in particular. t-SNE squeezes each company's many-dimensional footprint onto a page so that similar footprints land near each other. Neighbors are meaningful; the distance between two far-apart groups, and the left/right or up/down direction, are not.

A cluster

A group of companies whose conference footprints are alike: they sponsor and speak at the same conferences. Each is named after the conference that most set it apart. Clusters are computed in the full space, not from the picture, so a cluster can appear split on the map.

Method

Each company is a vector over the conferences where it sponsored (weight 2 per event) or spoke (1 per event), log-damped and TF-IDF weighted so that the ubiquitous flagships count less. The vectors are reduced with truncated SVD and grouped with k-means, picking the number of clusters by silhouette score; lookalikes are the nearest companies by cosine similarity. Only companies present at four or more conferences are included.

Clusters

18 clusters. Marker color repeats every 8 clusters, so the shape (dot, ring, diamond) and number tell them apart. Hover a card to light it up on the map.

0NeurIPS · ICML circuit138

ConferencesNeurIPS, ICML, ICLR, CVPR, ACL

e.g.Meta, Netflix, Uber, Apple, JPMorgan Chase

1MIT CDOIQ Symposium · Data Governance & Information Quality Conference circuit102

ConferencesMIT CDOIQ Symposium, DGIQ, Enterprise Data World, AWS re:Invent, CDAO (Chief Data & Analytics Officer) Summits

e.g.Walmart, Thoughtworks, KPMG, Roche, Mastercard

2Snowflake Summit · Coalesce circuit187

ConferencesSnowflake Summit, Coalesce, Databricks Data + AI World Tour, Snowflake World Tour, Data + AI Summit

e.g.Salesforce, Accenture, Deloitte, dbt Labs, Capgemini

3PGDays & regional Postgres events · PostgreSQL Conference Europe circuit78

ConferencesPGDays & regional Postgres events, PostgreSQL Conference Europe, PGConf.dev, PGConf NYC, Postgres Conference

e.g.Aiven, EDB, Zalando, Yugabyte, Fujitsu