The Everything Feed - All Packet Pushers Pods HN792: Understanding Agentic AI for Network Operations (Sponsored)
11 snips
Aug 15, 2025 Joining the discussion is Bob Friday, Chief AI Officer at Juniper Networks, now HPE, known for his expertise in AI and self-driving networks. The talk dives into the fascinating world of agentic AI and its role in network operations, emphasizing its nonlinear functionalities and the significance of high-quality data. Bob explains how these AI agents can autonomously interface with APIs and manage complex tasks, likening trust in AI systems to that of self-driving cars. Other highlights include the evolution of IT roles, the challenges of large language models, and the important balance of guardrails in AI operations.
AI Snips
Chapters
Transcript
Episode notes
Prepare APIs And Docs For Agents
- Start by cataloging where your data lives and ensure APIs and docs clearly describe those sources for agent consumption.
- Prepare API descriptions (OpenAPI/MCP) and improve docs so agents can reliably call functions without hallucinating.
Mix Big And Small Models Based On Needs
- Agentic systems combine large foundational models for reasoning with smaller specialized models to manage cost, latency, and privacy.
- Choose model types based on task complexity, cost, and data sensitivity rather than assuming one model fits all.
Use RAG Instead Of Training Models From Scratch
- Use RAG (embeddings + similarity search) to surface document answers without training large models from scratch.
- Chunk docs, create embeddings, and feed relevant chunks to the LLM for accurate, up-to-date responses.