Evidently AI Platform

Evidently AI develops an open-source framework (Apache 2.0) for evaluating and monitoring LLM applications, RAG systems, AI agents and predictive ML models. The site does not state the founding year or headquarters; the blog names Elena Samuylova as co-founder and CEO and Emeli Dral as co-founder and CTO.

The library runs automated evaluations with 100+ built-in metrics and LLM judges, generates synthetic test data, and tracks drift and regressions on a live dashboard. Release 0.7.17 added open-source LLM tracing and dataset management. The site cites users such as DeepL, Wise, Plaid and Realtor.com.

• LLM evals: 100+ built-in metrics plus custom rules, classifiers and LLM judges.
• Synthetic data: generation of edge-case and adversarial inputs.
• Monitoring: live dashboard for drift, regressions and emerging risks.
• Detection checks: hallucination, PII, toxicity and retrieval quality.
• Tracing: open-source LLM trace storage and viewer built on OpenTelemetry.
• ML monitoring: data drift and predictive performance tracking.

A Vietnamese bank can test its customer-support RAG assistant for hallucination and PII leaks before launch. A retailer can monitor demand-forecast model drift alongside a product-search chatbot. A software house can add automated evals to its CI pipeline.

An Apache 2.0 licence lets teams adopt evaluation tooling without licence cost. One framework covers LLM, RAG, agent and classic ML monitoring. Built-in checks for hallucination, PII and retrieval quality shorten test design. Synthetic edge-case data helps find failures before users do.

Teams shipping RAG and agent products need systematic evaluation instead of manual spot checks. Open-source evaluation libraries let regulated or cost-sensitive buyers keep data in their own environment. Vietnamese banks and telcos already running ML models can extend monitoring to generative AI.

Open source under Apache 2.0 with GitHub and Discord community; the site states use in thousands of companies. Co-founders Elena Samuylova (CEO) and Emeli Dral (CTO) named on the blog. Headquarters not stated.