Building Agents for Production: Lessons from a LangChain SRE Agent in Production

AI agents are easy to prototype but notoriously difficult to harden for production. This talk deconstructs the architecture of an autonomous SRE agent built on LangGraph and Deep Agents. We will explore how to manage complex, non-deterministic workflows by using narrow subagents, structural read/write splits, and proactive monitoring with human-in-the-loop safety. Learn the best practices for building agents for production and how to scale for enterprise usability and governance.

Key Takeaways:

1. Architectural Safety: Implement structural separation between autonomous read operations and human-gated write operations.
2. Modular Reasoning: Use narrow subagents for complex tasks to increase parallelism and reduce hallucination surface area.
3. Building for Scale: How to optimize agent performance for request and throughput scaling
4. Production-Ready Economics: Optimize LLM costs by routing tasks to specialized models (Haiku vs. Sonnet vs Opus) based on reasoning complexity.
5. Autonomous Agent Improvement: Using long term memory to improve performance and automation over time