Case Study: Snorkel AI achieves 2x faster issue detection with Portkey

Snorkel AI cuts agent debugging time in half with Portkey

Snorkel AI, a provider of expert data annotation for AI research, faced significant challenges debugging its complex multi-agent evaluation systems. Their fragmented workflow, involving scattered logs and manual reconstruction, made it difficult to trace failures within hierarchical agent executions. They turned to Portkey's observability platform to gain visibility into these multi-agent workflows.

By implementing Portkey, Snorkel AI gained a visual hierarchy that showed every agent call and tool invocation as a nested tree, enabling precise inspection of any prompt and response. This solution delivered substantial results: the team achieved a 20% accuracy boost in agent evaluations and cut issue detection time in half, moving from hours of manual log reconstruction to minutes of visual inspection. Portkey provided the critical visibility needed to debug with confidence.

Snorkel AI