AI Dev 25 x NYC | Nyah Macklin: How to Structure Context to Make Your Agents Smarter

2025-12-08 · Source: DeepLearningAI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Software Development & Engineering · Depth: Intermediate, long

Summary

Nia Mlin, a senior engineer and developer advocate for AI, presents on "architecting context" to enhance AI agent intelligence, focusing on structuring knowledge. The presentation addresses real-world problems like incident response, where generic AI agent advice is insufficient. Mlin introduces context engineering as a discipline for systematically providing models with relevant information, tools, and instructions in the correct format and time. Key techniques for improving context relevance include RAG plus hybrid search, memory management, context structure and ordering, and tools/function calling. The core focus is on knowledge graph augmented context, which encodes facts within relationships, enabling multi-hop reasoning and more accurate, traceable responses compared to traditional vector search methods. This approach has demonstrated significant improvements in accuracy and latency across industry benchmarks, such as an 18% accuracy increase and 90% latency reduction for AI agent memory with Zeppai, and LinkedIn cutting ticket resolution from 40 to 15 hours.

Key takeaway

For AI Engineers building agentic systems, prioritizing context engineering with knowledge graphs is crucial. Your agents will achieve higher accuracy and better explainability by leveraging structured relational knowledge, moving beyond generic responses. Implement knowledge graph augmented context to enable multi-hop reasoning and reduce reliance on less precise semantic search, significantly improving performance in critical applications like incident response or complex query answering.

Key insights

Knowledge graphs provide structured, relational context for AI agents, enabling multi-hop reasoning and superior accuracy.

Principles

• Context engineering builds dynamic systems for LLM invocation.
• Explicitly modeling relationships enables AI agent reasoning.
• Graph context consistently outperforms traditional RAG.

Method

Transform tabular data into a knowledge graph, then integrate it into an agent architecture using an MCP (Model Context Protocol) layer to provide structured, relational context for LLM reasoning and retrieval.

In practice

• Combine vector semantics with symbolic filters for hybrid search.
• Offload complex calculations to tools via function calling.
• Use knowledge graphs for traceable risk profiles and precise medical treatments.

Topics

• Context Engineering
• Knowledge Graphs
• AI Agents
• Retrieval-Augmented Generation
• Graph Algorithms

Best for: AI Engineer, Machine Learning Engineer, AI Architect

Related on AIssential

• How RAG, GraphRAG, and Context Engineering Improve AI Performance — Effective AI performance hinges on context engineering, not just raw model intelligence.
• Connecting the Dots with Context Graphs — Stephen Chin, Neo4j — Context graphs integrate LLMs with knowledge graphs to provide grounded, explainable, and auditable AI-driven decision-making by connecting disparate enterprise data.
• I Built an AI Data Engineering Mentor from Scratch: Here’s What Nobody Tells You — AI agent effectiveness is primarily driven by structured context, not just the underlying large language model.
• Stop babysitting your agents... — Brandon Waselnuk, Unblocked — AI agents require dynamic, comprehensive organizational context, not just intelligence or access, to operate effectively and produce mergeable outputs.
• What Is a Context Graph — and Why Is Everyone Talking About It? — Context graphs provide machine-readable institutional memory, enabling AI to accurately interpret complex organizational knowledge and decision logic.
• Beyond Vector Similarity: A Structural Analysis of Graph-Augmented Retrieval for Industrial Knowledge Graphs — Effective LLM-based graph reasoning hinges on providing robust computational operators, not just enhanced model intelligence.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by DeepLearningAI.