platoseedplatoseed
Open-source monitoring for machine learning models
We are building an open-source standard to monitor ML models in production. The tool is used by enterprise data science teams to operate their models reliably and detect and resolve issues.
Evidently AI provides an open-source evaluation and observability platform for AI/ML systems, focusing on evaluating model quality, safety, and data drift. It offers automated evaluation, synthetic data generation, continuous testing, and a library of metrics to monitor production AI systems.
The platform enables automated evaluation of AI outputs with reports that pinpoint where AI breaks, supports synthetic data creation for edge cases and adversarial inputs, and provides continuous testing with a live dashboard to track drift and regressions across updates. It includes a library of 100+ in-built metrics, the ability to add custom metrics and rules, and supports adversarial testing, RAG evaluation, and multi-step AI agent workflows. It is built on top of an open-source AI evaluation foundation and emphasizes transparency and extensibility with its open-source origins and GitHub ecosystem.
Who it’s for: Teams building and maintaining AI/ML systems who need rigorous testing, observability, and governance for model quality, safety, data drift, and retrieval quality in production.
Community-driven with GitHub presence, open-source tooling, and a large ecosystem (stars/downloads/community).
- CEO and Co-founder Evidently AI - Spent 7 years helping large enterprises deploy ML systems - Previously built an industrial AI startup focused on steel and chemical markets - Led business development for an enterprise ML division at Yandex (NASDAQ:YNDX)
- CTO and Co-founder Evidently AI - Co-author of the most popular Coursera course on ML in the Russian language with 100K+ students - 50+ real-world ML projects from e-commerce to manufacturing - Ex-Chief Data Scientist at Yandex Data Factory, enterprise ML division at Yandex
The fastest way to ship airtight AI
Automated data observability and quality for Data Lake
