DataRobot launched new AI observability functionality with real-time intervention for generative AI solutions, available across all environments including cloud, on-premise and hybrid.

This latest release provides AI leaders and teams with the tools to confidently build enterprise-grade applications, manage risk and deliver business results.

“Lack of visibility and risk are significant obstacles to reaching real business value from AI,” said Venky Veeraraghavan, Chief Product Officer, DataRobot. “We're revolutionizing AI observability with real-time intervention across diverse AI assets and environments, so leaders can safeguard projects, up-level oversight and empower teams."

This announcement brings AI observability for any AI asset and environment into the DataRobot AI Platform to deliver:

- Cross-Environment AI Observability: Gain full oversight across environments and reduce risk across your entire AI landscape with unified governance for all predictive and generative AI assets.

- Real-Time Generative AI Intervention and Moderation: Build a multilayered defense to safeguard AI applications with customized build, intervention and moderation workflows, leveraging a rich library of pre-built and configurable guards to ensure accuracy and prevent issues like prompt injections and toxicity, detect personally identifiable information (PII) and mitigate hallucinations.

- Generative AI Alerts and Diagnostics: Gain control and flexibility with customizable alert and notification policies, visually troubleshoot problems and traceback answers, and set robust multi-language diagnostics with insights for data quality checks, topic drift and more.

This new release also introduces best-in-class evaluation, testing and open source LLM support capabilities:

- Enterprise-Grade Open Source LLM Hosting: Leverage any open source foundational model including LLaMa, Hugging Face, Falcon and Mistral with DataRobot’s built-in LLM security and resources, complementing recent integrations with NVIDIA NIM inference microservices and NVIDIA NeMo Guardrails software to accelerate AI deployments for enterprises.

- LLM Evaluations, Testing and Metrics: Enhance application quality, assess LLM performance and automate testing with groundbreaking out-of-the-box synthetic test data creation, evaluation metrics and quality benchmarks.

- Advanced RAG Experimentation: Evaluate different embedding methods, chunking strategies, and vector databases to assess and identify the best RAG strategy for each use case.

All of the functionality announced today is available on cloud, on-premise, and hybrid environments.