Founded in 2024, Isaeus is the brainchild and passion project of its lawyer/engineer founder. We are currently building our models and hope to begin offering them to clients in 2026. Our genesis reflects our passion for the legal profession and our commitment to facilitating the crucial, arduous work lawyers perform.
Isaeus plans to offer paid access to our Promethean, domain-specific large language models ("LLMs"). Isaeus will not offer a virtual lawyer. Rather, we aspire to aid lawyers by providing an intelligent, automated assistant to generate templates, flag drafting or grammatical errors, and offer suggestions.
Our founder is a machine learning engineer; software engineer; and active lawyer. An honors graduate of an elite US law school, he is a self-professed "math nerd" with deep knowledge of natural language processing, multivariable calculus, linear algebra, and mathematical stats/probability.
Unlike many participants in the natural language processing space, Isaeus will not attempt to reduce our entire, incomprehensibly complex world to an ostensibly simple set of 1s and 0s. Rather, we limit our models' knowledge to the legal milieu; while immense complexity is inherent in all LLMs, eschewing disparate domains extracts as much noise as possible. We hypothesize this epistemological reduction will engender a more specialized agent better able to navigate the legal world's semantic minutiae and avoid the hallucinations that plague other, general LLMs. The transformer architecture underpinning LLMs is novel—it originated in a 2017 Google paper—and researchers worldwide, including us, learn more each day. Those who endeavor to merely avoid failure inevitably find it. Thankfully, Isaeus harbors grander ambitions irreducible to a success/failure dichotomy: if our hypothesis holds, our models will secure their place in the vanguard of domain-specific LLMs. Regardless of the outcome, we are happy to help advance this nascent field's epistemological frontier.