Shawkat Kabbara is the founder and CEO of Papr, the context intelligence layer for AI. Shawkat has spent most of his career in search and machine learning with roles at Microsoft (Bing), Yahoo, and Meta (Facebook app search) and most recently at Apple, where he was the product founder of the App Intents Swift SDK powering Apple Intelligence Actions, helped launch Vision Pro, and optimized ML models to run on memory-constrained devices. He started Papr to solve a problem he kept hitting himself: AI systems can find similar content, but they can't reason over the relationships that actually drive business decisions. Today Papr's graph-native memory, built on Neo4j, powers AI agents for companies in regulated industries where provenance and auditability aren't optional.

Papr.ai is a context intelligence platform that gives AI agents long-term memory through a graph-native architecture built on Neo4j. Papr powers memory and retrieval for customers in fintech, commerce, healthcare, and insurance, turning unstructured data into queryable knowledge graphs via developer-friendly memory policies and a GraphQL API.

In this episode Shawkat walks through a before-and-after: 164 daily sales reports stuck as flat text, then the same data as a commerce graph an operator can query in real time.
The core question every guest answers: *describe the moment you realized a graph was the right tool.*
If your stack is straining against the questions you actually want to ask, this one's for you.

• What "context intelligence" means for AI agents
• Why flat text fails at scale for memory and retrieval
• Building a graph-native memory layer on Neo4j
• Memory policies and the GraphQL API developer experience
• The before-and-after: 164 sales reports → a queryable commerce graph
• Serving fintech, commerce, healthcare, and insurance use case