Ryan Hynes

Ryan Hynes is a practitioner and researcher working at the intersection of data, law and technology. Mr. Hynes builds tools and datasets to help clients meet their e-discovery needs and embrace new technologies. Bringing more than 10 years of experience in the e-discovery field, Mr. Hynes leads data science initiatives and rigorous testing and evaluation of large language models across cloud platforms. He also supports FTI Technology teams and clients with developing benchmark datasets to assess generative artificial intelligence capabilities for a variety of e-discovery tasks.

Mr. Hynes is a U.S.-qualified attorney, software engineer and data scientist working closely with clients to deploy and fine-tune generative AI solutions to meet specific use cases. Leading numerous matters involving large language models, vector search, retrieval-augmented generation and information retrieval, Mr. Hynes’ years of experience allow him to understand and apply both existing and emerging machine learning techniques and models to e-discovery challenges. Mr. Hynes was part of the team responsible for building IQ.AI by FTI Technology™, an advanced suite of AI-driven offerings designed to address data-intensive challenges in legal and compliance matters.

In addition, Mr. Hynes is experienced in model evaluation and validation, a growing concern with the rise of non-deterministic models. He maintains an active research profile, regularly publishing and presenting his work in academic journals and at conferences, contributing further to advancements in his fields.

Prior to joining FTI Technology, Mr. Hynes served as in-house counsel and software architect at Relativity, where he worked on predictive coding, classification and search. Notably, Mr. Hynes spearheaded the creation of Relativity Patents, a semantic prior art search engine for patent attorneys, leveraging an expansive eight billion record vector search index of global patents.

Relevant Experience:

• Developed e-discovery-specific benchmark datasets and validation processes for classification, summarization and extraction, utilising human reviewers from FTI Technology’s managed document review team to create labeled, ground truth data
• Created an experimental large-language-model-based privilege review workflow, combining traditional machine learning classifiers and numerous models to automate privilege review and privilege log generation
• Designed an answer evaluation framework for a large-language-model-based legal research assistant; the metrics output from this framework were used to further refine prompt strategies and led to dramatically improved answer quality