Automorphic

What is Automorphic?

Automorphic is a platform that enhances language models with knowledge infusion, data conversion, and security tools. It consists of three core components — Conduit, TREX, and Aegis — plus the Automorphic Hub, and integrates with the OpenAI API so you can use these features without changing existing code.

Conduit

Conduit infuses knowledge into language models. It supports fine-tuning, lets you combine behavior or knowledge adapters on the fly, and enables fast iteration with human feedback.

TREX

TREX converts unstructured data into the formats you define, including JSON, XML, YAML, or custom types specified with regular expressions or context-free grammars.

Aegis

Aegis acts as a firewall for language models, detecting and defending against adversarial attacks, personal information (PII) leaks, and toxic language. It improves over time as it learns from usage.

The Automorphic Hub provides access to models trained and improved on the platform.

Who created Automorphic?

Automorphic was developed by an organization focused on protecting models and users, specifically guarding against adversarial attacks, prompt and PII leakage, and toxic language. Specific founder and company details are not available.

Who is Automorphic for?

Automorphic serves data scientists, machine learning engineers, AI researchers, software developers, data engineers, data analysts, and other AI practitioners.