Design and Implementation of the RelationalAI Knowledge Graph Management System
summary
RelationalAI implements a cloud‑native relational knowledge‑graph engine that keeps the immutable versioned catalog for O(1) clones and strict serializability, evaluates thousands of interdependent relations on demand with Salsa/Differential Dataflow, and exposes a composable rel language for libraries, relational ML (no denormalized design matrix), and solver specs. Query execution combines a Julia compiler with a vectorized interpreter and leverages worst‑case optimal joins and a new dovetail join for correlated patterns. Customers see ~10x application code reduction; system pairs with DuckDB and object storage for terabyte+ data.
AI-written summary · about
chapters (19)
- 0:00Start
- 3:49Modern Databases Are Cloud Native
- 10:14Label Property Graphs
- 12:54Matrix Multiplication
- 27:07Object Storage
- 32:36Join Algorithms
- 32:48Binary Join Algorithms
- 34:04Optimal Joint Algorithms
- 34:46Sparsity
- 38:33The Nested Loop
- 39:01Semantic Optimizer
- 44:39Doctoral Drone Compiler
- 45:29Semantic Optimization
- 52:10Integrity Constraints
- 52:46Relational Machine Learning
- 55:58Mathematical Optimization
- 56:22Constraint Optimization
- 58:27Sql
- 59:05Query Evaluation
related talks
14mIntroducing RelationalAI
44mBuilding Knowledge Graphs- 44mBuilding Knowledge Graphs
46mBuilding Knowledge Graphs
24mWhy You Need a Knowledge Graph And How to Build It
41mMerging the relational, the graph and the semantic
15mGraphRAG-Powered AI Agents: Transforming Corporate Metadata into Intelligent Knowledge
21mBringing GraphQL Natively To Relational Databases With AI