Galileo AI

Galileo is a closed-source AI evaluation and observability platform founded in 2021 by Vikram Chatterji, Yash Sheth, and Atindriyo Sanyal, who previously built AI systems at Google AI, Google Brain, and Uber AI respectively. The platform is now part of Cisco, following a completed acquisition on May 22, 2026, and is being integrated into Splunk Observability Cloud. Its core technical differentiator is Galileo Luna, a family of proprietary Evaluation Foundation Models trained specifically for evaluation tasks rather than general language generation, which Galileo argues produces faster and more accurate hallucination detection than prompting a general-purpose LLM to evaluate outputs.

Key Capabilities

• Luna Evaluation Foundation Models (EFMs): Purpose-built evaluation models fine-tuned on task-specific datasets for hallucination detection, groundedness scoring, and factuality measurement, operating as a proprietary alternative to LLM-as-judge approaches

• Agentic evaluations: Full lifecycle tracing for multi-step agents with step-by-step error detection, tool call analysis, and system-level performance metrics across planning, execution, and completion stages

• RAG evaluation metrics: Specific measurements for context adherence, retrieval completeness, and knowledge base coverage across retrieval-augmented generation pipelines

• Production monitoring with guardrails: Real-time scoring of live requests with automated guardrail enforcement and alert-based detection of systemic failures including misaligned tool calls and cost or latency regressions

• Continuous learning with human feedback (CHLF): A feedback loop that routes low-scoring production outputs back into evaluation datasets, enabling iterative improvement grounded in real user interactions

• Splunk Observability Cloud integration: Post-acquisition, Galileo extends Splunk's AI Agent Monitoring capabilities, consolidating agent behavior telemetry with existing network and security observability data 

See Galileo AI Pricing Details →