R&D Innovations Repository

Upgrading core software without breaking old models.

Training an LLM to detect lookalike domains to thwart attempted vendor fraud.

Using LLMs to produce embeddings that represent semantic meaning of a word or phrase in order to categorize text.

Using LLMs for large scale conversation evaluation.

Building ChatGPT for Security Analysts

Counting the frequency of certain suspiciousness characteristics in a customer's environment helps improve the performance of our detection engine.

The detection team utilizes a large language model to automatically assign names to complex detection rules.

We've built a large language model that can interpret our detection signals to make an attack/safe judgment on a message.

A large language model generates detailed descriptions of each signal that our detection engine utilizes.