Graphistry, Inc.

How Graphistry won the US Cyber Command AI competition by automatically correlating alerts into incidents and kill chains. Many teams already like to use Graphistry to quickly cluster alerts like network scans because of its quick graph ML and visualization flows. The US Cyber Command AI competition needed us to then zoom out and automatically explain the broader incident timeline. The clusters got us half way there: By linking the relationships between the clusters, we can see an incident as a graph-of-graphs, and quickly build all our analyses, reports, and visualizations on top! In this video, Leo Meyerovich (Graphistry CEO and co-founder) demonstrates the flow our small competition team built over a week using just one Graphistry GPU node to tame a billion+ alerts from a real cyber fusion center: ? Event AI: Using UMAP to fuzzy match similar events into clusters. ? Event hypergraphs: Linking clusters to show how events progress and influence. ? Event graph classification: Assigning cluster IDs and incident IDs to track distinct behaviors, then using direct relationships to map attack progression. Example: When analyzing network requests, UMAP reveals not just events in the same scan, but different phases of the scan attempts and subsequent anomalous behavior on the device — critical understanding for scoping, root cause analysis, and understanding the progression. Key steps: ? Cluster events by similarity using scalable autoML, e.g., group scans via GPU UMAP. ? Analyze linkages within clusters to surface progressions, e.g., via hypergraphs. ? Map relationships to phases like reconnaissance, exploitation, etc. by labeling clusters via graph ML and then summarizing each cluster with their ATT&CK labels. ?? Join the discussion Connect with us in the Graphistry community Slack. ??Check out the full NDC conference talk find the link in the comments.

UMAP: The first tool we reach for to instantly make sense of almost any 1,000+ row dataset Clustering rich data doesn’t have to feel like deciphering a technical manual. UMAP clustering is an easy way to simplify the chaos. In this video, Leo Meyerovich (Graphistry’s CEO and co-founder) shows an example of breaking down UMAP’s power into plain language: ??Cluster columns of data, whether numeric, IDs, dates, text, or a rich combination, with a simple automated approach that outperforms older methods like PCA ??Transform wide tables – think dozens and even hundreds of columns! – into intuitive visual clusters ??Leverage the last 5 years of innovation: UMAP handles modern datasets faster and more accurately than legacy techniques through automatic GPU acceleration and optional automatic model learning. Whether you’re in Jupyter, Colab or Databricks notebooks, UMAP unlocks patterns hidden in rows and columns with just a few lines of code and you can easily visualize the clusters and similarity using Graphistry. Video highlights: ?? One-liner clustering for non-linear, high-dimensional data. ??Go beyond PCA/t-SNE with faster, more flexible dimensionality reduction. ??Works out-of-the-box on large-scale datasets that trip up traditional tools. ?? UMAP starter notebooks in the first comment ?? Join the conversation comment “UMAP” to receive an invite to the Graphistry community slack

Unlock the power of graphs—comment “Viz” to secure your webinar spot now!! #GraphAI #DataScience #BigData #MachineLearning #GraphAnalytics #Python #EnterpriseAI

Exciting News: Graphistry + Google Spanner Graph Enterprise teams struggle to analyze massive, connected datasets in real time—but not anymore. Graphistry and Google Spanner Graph are teaming up to revolutionize real-time, multi-modal graph analytics at scale. Here’s what this powerhouse duo delivers: ?? Instant Insights: Transform massive datasets into actionable insights in seconds—no more getting stuck in endless tables. ?? Smarter Investigations: Detect fraud, neutralize cyber threats, and optimize operations with AI-powered graph visualizations. ?? Multi-Modal Powerhouse: Unlock relational, full-text, vector similarity, and graph pattern matching capabilities—all seamlessly integrated with SQL and GQL for comprehensive analysis. From cybersecurity to fraud detection to customer intelligence, this partnership empowers teams with next-gen analytics built for speed, scale, and precision. Stay tuned for real-world use cases and hands-on demos. #GraphAI #DataScience #BigData #MachineLearning #GraphAnalytics #Python #EnterpriseAI

Graphs: The Key to Unlocking Advanced Data Analysis Comment "Data" to be invited to our upcoming webinar. #cybersecurity #security #technology #hacking #hacker #infosec #ethicalhacking #cybercrime #cybersecurityawareness #hackers #cyberattack #informationsecurity #datasecurity #dataprotection #malware #programming #coding #hack #privacy #ethicalhacker #computerscience #cybersecuritytraining #graph

Wondering how to turn a table of 100,000+ data points from your Jupyter and Databricks notebooks into interactive graphs? Turning tabular data into graphs reveals insights that are hard to see with rows and columns, and it’s simpler than you might think. Leo Meyerovich, Graphistry founder, shares 3 popular one-liners PyGraphistry users use to explore the tables and dataframes as graphs. Video highlights: ?? Convert tabular data into interactive graph visualizations with just one line of code. ?? Use AI-powered clustering to surface hidden patterns automatically. ?? Build hypergraphs for more control in connecting multiple dimensions in your data. ?? Quickly use #Graphistry to visualize graphs in Jupyter, Colab, and Databricks with many features out-of-the-box, even on datasets too big for regular graph tools. ?? Find the link to our sample notebooks in the first comment! ?? Comment “Security” to be invited to our upcoming webinar! #cybersecuritytips #CyberCrime #cybersecurity #datascience

Curious how to bring genAI to your security pipelines and scale to 1 billion+ data points —across OSINT, logs, and alerts —all in real-time? Leo Meyerovich, co-founder of Graphistry, spoke at SuriCon 2024 with Anthony G. Tellez (BNY), including live demos of genAI-powered rule and log management that build on best practices for building GPU, graph, and AI pipelines. Key highlights: ?? LLMs for security case study: AI Rule management and AI threat intelligence ?? Building and scaling RAG pipelines ?? Smarter RAG through Graphs: Multi-hop reasoning with PyGraphistry/GFQL ?? Embeddings crash course: How to think about them, how to scale them ?? Using Louie.AI to build security genAI workflows like these for 100X efficiency Want the slides and the full YouTube talk? ?? Find the link in the first Comment!

Proud to announce the Graphistry 2025 GPU AI Remote Fellowship program! We're looking for 2 students (PhD/MS/BS) to hack on autonomous AI analysts and multi-gpu OSS graph query languages. These projects in turn power how we're working with banks, NGOs, the US gov, and others to tackle core problems in cybersecurity, disaster response, anti-misinformation, and a surprising number of other society-level applications. For PhD-level students, the results can also lead to publications. On a personal note, one of the reasons we started Graphistry was to build an organization who could grow to support this kind of meaningful research, so we're quite excited about the coming months! Please share with ambitious students you know and may be looking for the right opportunity :)

Curious about getting started with graphs and ML+AI, and then how to scale it? Great class at Connected Data this week, including hands-on time with Graphistry, Inc. GPU visual graph analytics for your notebooks and pipelines. RSVP info below. #ConnectedDataLondon #GraphML #GNN

#proudpartner ! cugraph is one of the powerful GPU components in how our open source GFQL graph query language is in turn making folks able to get graph500 results. Our internal graph500 benchmarks 11 months ago, on cheapo $0.25/hr GPUs, had it around *#100*, so I am not surprised to see a more real setup hit top #3, and am looking forward to the next rig's results. Congrats Joe Eaton Bradley Rees & and the #GPU #RAPIDS community! cc Brad Spiers