Mori

Mori was a genealogy platform with the mission of helping people expand their family trees. A surprising amount of genealogy research is still done manually, and Mori offered ways to extract clues out of a user's existing family tree and assist in the search for new relatives.

I was brought in to complete the multiplayer and commenting features on the main family tree builder, before working on a conversational interface for providing research suggestions to users.

The average expert in an industry is not necessarily an LLM-prompting expert. In late 2023, our researched showed that while LLM models were capable of extracting and summarizing data from a GEDCOM (genealogical data format) file, getting actionable insights required understanding both what was on the GEDCOM, and also what the model was capable of.

Our guardrails prevented users from freely prompting the model, rather, it would present to users who the least-researched relatives are, and suggest that as a starting point. This takes a complex file format and distills it down into a clear summary along with clues for further research.

The release of the Pinston was met with positive reception and increased the amount of family trees uploaded to Mori by 3x in a period of a month. Users remarked that being able to see gaps in their family tree written out in plain English helped with figuring out where to search next for more information about their relatives.