See the network hiding in your tables.
Agentic Graph Data Scientist. Bring one table of payments, customers, purchases or suppliers. In one session we turn it into a Neo4j graph, show what the connections reveal, and hand you the code to load it.
Talk to the founder,
Aaron Tekle
Generated code is never run against your database. You review it and run it yourself.
| sender_id | receiver_id | amount |
|---|---|---|
| acct_09 | acct_31 | 7,620 |
| acct_31 | acct_44 | 7,480 |
| acct_44 | acct_09 | 7,350 |
| acct_12 | acct_31 | 1,950 |
| acct_27 | acct_12 | 4,970 |
Where graphs pay off
If your data records who paid whom, who shares what, or what depends on what, the answer you need is probably in the connections, not the rows.
-
(:Account)(:Account)
Fraud rings and mule networks
Find accounts that pass money in loops, fan out to many new receivers, or sit at the center of a tightly connected group.
-
(:Customer)(:Device)
Customer 360 and entity resolution
Spot records that are likely the same person because they share a device, an address or a card, even when names are spelled differently.
-
(:Customer)(:Product)
Recommendations
Recommend from what similar customers bought, using node similarity instead of hand-maintained rules.
-
(:Supplier)(:Part)
Supply chain risk
See which suppliers many products quietly depend on, and what stops if one of them does.
How a session runs
Cleo does the modeling, analysis and code generation live. We walk through every step with you, so the model matches how your business actually works.
Share a sample
One table as CSV, Parquet, JSON or Excel. A masked extract of a few thousand rows is plenty.
Model it together
Cleo proposes node labels, a relationship type and keys. We adjust it with you until it reads the way your team talks.
Read the signals
PageRank for influence, Louvain for communities, connected components for clusters, and what each can and can't tell you.
Take the package
You leave with Cypher, a Python loader and a Graph Data Science workflow to run in your own Neo4j.
What you leave with
- A graph model with node labels, relationship types, keys and the properties worth keeping.
- Neo4j code: constraints, parameterized load statements and a Python ingestion script.
- A Graph Data Science workflow that projects the graph and runs PageRank, Louvain and connected components.
- A short findings summary covering the signals we found, what they suggest, and what to check next.
neo4j_package.zipExampleREADME.mdHow to run it01_constraints_and_load.cypherCypher02_ingest.pyPython driver03_gds_workflow.cypherGraph Data Scienceschema_mapping.jsonThe model
// 01_constraints_and_load.cypher CREATE CONSTRAINT IF NOT EXISTS FOR (a:Account) REQUIRE a.id IS UNIQUE; UNWIND $rows AS row MERGE (s:Account {id: row.sender_id}) MERGE (t:Account {id: row.receiver_id}) MERGE (s)-[r:SENT]->(t) SET r.amount = row.amount;
Pick a format
Start small. Most teams begin with a discovery call and move to a working session once the data question is clear.
| Detail | Discovery call | Working session | Pilot |
|---|---|---|---|
| Length | 30 minutes | Half a day | Two weeks |
| You bring | A description of your data and the question you want answered | One sample table | A masked extract and the people who know the data |
| You leave with | A recommended graph model and whether a graph is the right tool | The model, first findings and the full Neo4j package | A package tuned on your data and a findings review with your team |
Expected (Common)Questions
Do we need Neo4j already?
No. The package includes constraints and load scripts you can run on your own Neo4j instance or a managed one such as AuraDB.
What data works best?
A table where each row links two things: a payment between accounts, a purchase by a customer, a part from a supplier. Extra columns become properties or edge weights.
Who is Cleo?
Cleo is an AI agent for graph data science. It drafts the model, the analysis and the code, and we review all of it with you in the session.