Grounded Reasoning Systems for Cloud Architecture - Iman Makaremi
AI Summary
In this talk, Iman Makaremi explores the architectural challenges faced by developers as LLMs integrate into enterprise workflows. It details the design and implementation of an AI orchestration framework that leverages LangGraph for multi-agent workflows, Flyte for distributed execution, and AWS Bedrock for LLM inference using Claude 3. The system includes a knowledge graph serving as long-term memory for agents, enabling structured representations of architectural states and business contexts which help generate accurate recommendations and translate natural language into SQL or code. Additionally, the talk discusses integrating observability practices with OpenTelemetry to debug AI systems in production, providing a comprehensive blueprint for developers and AI engineers aiming to embed reasoning into complex system design.