AI BriefingHow Schneider Electric Built Its LLMOps Foundation with LangSmith
Key point
Schneider Electric leveraged LangSmith to build an LLMOps system for managing observability, evaluation, and deployment of large-scale AI agents.
Details
Global energy technology company Schneider Electric operates an AI Hub staffed by 350 experts and has deployed more than 60 AI agents to optimize energy consumption and boost productivity. They are focused on scaling AI while maintaining data security and quality in complex infrastructure environments.
Existing MLOps approaches had limitations when it came to debugging agent behavior in LLM systems, precisely measuring the impact of prompt changes, and verifying the production readiness of GenAI systems. To address this, Schneider Electric organized its LLMOps capabilities around three core pillars: Observability, Evaluation, and Deployment.
On the observability side in particular, the company built a Self-hosted deployment of LangSmith on AWS EKS to strengthen data privacy. It also runs a single workspace per product to manage traces end-to-end from Dev to Prod, creating a structure where production data is instantly converted into datasets to continuously improve agent performance.
Through this system, the company tracks all conversations from 'One Jo', an internal AI assistant serving 160,000 employees across 107 countries, and builds regression test datasets based on real-world use cases to increase the reliability of the system.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.