Highlights:

  • The Patronus platform is acclaimed as the inaugural automated evaluation and security platform that assists companies in safely utilizing LLMs.
  • Patronus AI revolutionizes model evaluation in enterprises by automating and scaling manual, costly processes. This ensures confident deployment of LLMs, reducing risks of model failures and misaligned outputs.

Patronus AI Inc., an automated evaluation and security platform, recently emerged from stealth mode, revealing three million dollars in seed funding.

Established by machine learning specialists Anand Kannappan and Rebecca Qian, former employees of Meta Platforms Inc., Patronus aims to address the complexities of assessing AI outputs and evaluating large language models for businesses.

The Patronus platform is acclaimed as the inaugural automated evaluation and security platform that assists companies in safely using LLMs. Leveraging AI, the platform empowers enterprise development teams to evaluate model performance, create adversarial test cases, and benchmark models effectively.

Patronus AI revolutionizes model evaluation in enterprises by automating and scaling manual, costly processes. This ensures confident deployment of LLMs, reducing risks of model failures and misaligned outputs. Leveraging machine learning technology, the platform rigorously tests and scores language models, pinpointing potential failures for enhanced reliability.

This platform automates the scoring and ranking of model performance, evaluating real-world scenarios and crucial criteria such as hallucinations and safety. Patronus AI automatically generates expansive adversarial test suites at scale and conducts benchmarking to assist customers in identifying the most suitable model for their specific use cases.

Chief Executive Kannappan said, “Every company is looking for ways to use LLMs today, yet they are concerned that unexpected model behavior, incorrect outputs and hallucinations will put their business and customers at risk. Whether off-the-shelf, open-source or custom, models today remain inadequately vetted and tested in real-world scenarios. And until now, the process of evaluating LLMs has been extremely inefficient and unscalable, producing unreliable results.”

The company initiated operations with a USD 3 million seed funding round, prominently backed by Lightspeed Venture Partners L.P. Other significant participants encompassed Factorial Capital LLC, Amjad Masad (CEO of Replit Inc.), Gokul Rajaram, in addition to multiple executives and board members from Fortune 500 corporations.