From Theory To Deployment: Building Scalable Multi-Agent Question-Answering Syste...
Talkat Open Source Summit Japan + AI_dev: Open Source GenAI & ML Summit + Automotive Linux Summit 2025 · 2025-12-08..12-10·
summary
A production-ready blueprint that deploys a modular multi-agent RAG system on CPU, showing a supervisor agent orchestrates specialized worker agents that include a RAG worker + SQL worker (vector DB retrieval and SQL extraction) and microservices for ranking, data prep, telemetry and one‑click Helm deployments to AWS/GCP/Azure or on‑prem. Claims AMX integer8 (and promising int4) roughly doubles LLM throughput versus AVX, recommends AWQ quantization, and reports a 32‑concurrent test with low‑second response times.
AI-written summary · about
chapters (10)
- 0:00OPA project introduction
- 2:10Hypermodel vs domain models
- 5:59CPU acceleration for AI
- 7:29Open Platform for Enterprise AI architecture
- 12:32Enterprise use cases
- 23:04Defining AI agents
- 27:20Agent examples tutorial
- 31:00Multi-agent applications
- 33:06Enterprise inference endpoints
- 35:18Summary and future
related talks
24mAI Agent Swarms Are Amazing
13mLightning Talk: Scalable and Observable RAG for Question-Answering in a Box
23mBuilding Production Multi-Agent RAG Systems on Serverless AWS
30mLessons From Building and Deploying a Multi-Agent System at Scale: An...
45mFrom RAG to Enterprise AI Agents: Building Intelligent Java Apps
46mAI Agents for the Rest of Us
51mOptimizing Complex Workflows with Event-Driven Multi-Agentic Approach
29mBuilding Scalable AI Agents with Ray