Arize AI

Arize AI is an observability and evaluation platform for machine learning models and LLM applications running in production. It tracks things like data drift, feature drift, model performance degradation, and embedding quality, and gives teams dashboards and tracing tools to see why a model's predictions or an LLM's outputs are going wrong, rather than just watching an aggregate accuracy number slip. For LLM specifically, it covers tracing of multi-step agent and RAG pipelines, prompt and response evaluation, and hallucination or toxicity checks, building on Arize Phoenix, its open source tracing and eval library that a lot of teams adopt before moving to the paid platform.

It's aimed at ML engineers and data science teams who ship models or LLM-based products and need to know when something breaks after deployment, not just before it. Use cases range from catching silent drift in a fraud model to debugging why a customer support agent built on GPT or Claude started giving worse answers after a prompt change. The platform supports both traditional ML monitoring (tabular models, embeddings, CV) and the newer LLM/agent observability workflows, so teams don't need separate tools for each.

What sets it apart is that breadth combined with the open source Phoenix project, which has become a common entry point for developers instrumenting LLM apps with OpenTelemetry-based tracing before they need enterprise features. Arize competes with tools like WhyLabs, Fiddler, and Galileo on the ML/LLM observability side, and with LangSmith and Langfuse more specifically on LLM tracing, but it's one of the more established names in the space with a long track record in production ML monitoring predating the recent LLM wave.