Inside LinkedIn's cognitive memory agent for agentic personalization
Ryan is joined by Praveen Bodigutla, Principal AI Researcher at LinkedIn, to chat about the four-layer memory system his team built to give LinkedIn's hiring assistant a persistent, personalized state.
LinkedIn's cognitive memory agent for agentic personalization is a system designed to provide personalized experiences for recruiters. The project was initiated after LinkedIn's hiring agent and hiring assistance received feedback from recruiters expressing their hiring preferences and refining roles they were hiring for. LinkedIn's goal was to create a memory agent that provides a deep personalization layer for recruiters, managing the entire life cycle of memory from understanding the ingested information, to retrieving and contextualizing the data, updating the memory, and providing provenance.
The agent uses four layers of memory to achieve this goal: the conversation memory, which represents the most recent and up-to-date preferences; the semantic memory layer, which aggregates information from across different interactions and sessions; the episodic store, a temporal querying layer that provides specificity of signal and provenance; and the procedural memory layer, which captures patterns and sequences of user activities across different platforms and services.
Together, these layers allow for a more personalized, context-aware experience for LinkedIn's recruiters.
Written by urgent.news from Stack Overflow Blog's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
