DataOps AI Agent That Fixes Pipeline Failures
Client: US delivery-tech unicorn
A LangGraph multi-agent system that monitors Dagster, reads failure logs with an LLM and applies known fixes without paging an engineer.
The Challenge
Engineers were being woken up for pipeline failures that had well-known fixes, such as out-of-memory errors and stuck PostgreSQL replication slots.
Our Solution
Built a multi-agent system with LangGraph that monitors Dagster, uses an LLM to read failure logs, and applies the known fix: it scales compute and restarts the job, or for replication alerts it stops the job, removes the stuck slot and starts a partial refresh.
Outcomes
Common failures fixed automatically, without paging an engineer
Replication-slot incidents handled end to end
Agent actions visible in Grafana
Technologies
Have a Similar Project?
Tell us about your data or AI challenge and we'll show you how we would approach it.