Calvin Qi

Calvin Qi leads the Retrieval and Data team at Harvey, building the knowledge and search infrastructure behind its legal AI products. His work combines enterprise legal retrieval, expert-guided evaluation, and specialized models to help lawyers answer complex questions without compromising confidential client information.

Qi earned bachelor’s and master’s degrees at Stanford University. He worked on machine-learning systems optimization and recommender models at SambaNova Systems before joining Glean, where he developed machine-learning and natural-language-processing models for semantic search. At Harvey, he applies that search background to legal research spanning private documents, legislation, case law, and tax regulations.

• Search built for legal complexity. A single legal question can combine statutory identifiers, specialized terminology, date restrictions, jurisdiction, and multiple authorities. Qi’s approach to enterprise-grade retrieval combines semantic and keyword search with metadata filtering while accommodating large-scale ingestion, responsive queries, customer isolation, and document-retention requirements.

• Legal embeddings shaped by expertise. Qi contributed to Harvey’s collaboration with Voyage AI on custom legal embeddings, including voyage-law-2-harvey, which incorporates legal texts and expert-annotated examples to improve retrieval of relevant authorities and passages.

• Investigative, citation-grounded search. His work on agentic legal research describes systems that plan investigations, choose among knowledge sources, assess evidentiary gaps, and search again before producing citation-backed answers.

• Evaluation calibrated to the stakes. Qi advocates expert-guided evaluation that combines intensive professional review, specialist-defined assessment criteria, and automated retrieval precision and recall. This layered approach supports rapid experimentation while accounting for confidential customer data and legal nuance, priorities he outlined at AI Engineer World’s Fair 2025.

• Specialized models for sustained legal work. Qi coauthored the Harvey Tenet research preview, introducing Harvey Tenet, an open-weight model post-trained with Fireworks research for long-horizon legal tasks.