Consulting

Our consulting methods are drawn from qualitative research traditions used in human-computer interaction and policy research. A typical engagement includes semi-structured interviews with the people who handle the documents, records, or workflows in question, and may also include surveys, contextual inquiry, and on-site observation depending on the nature of the problem. Interview transcripts are coded thematically, and findings are organized into a written report that identifies patterns, names points of friction, and specifies the requirements a better system would need to meet.

Project Development

Our development work centers on document intelligence and the data infrastructure that supports it. When an archive begins on paper or other physical media, we handle the digitization step. We then build parsers that combine computer vision and layout analysis to extract structured information from the resulting files (hello query!), along with the pipelines that manage the inference and training runs behind these parsers. Every result can be traced back to the model, code, and data that produced it, which matters when the output will be cited in a brief, a publication, or a filing. When an engagement calls for an AI component, solutions can be developed and run on dedicated local hardware, or within your own infrastructure when that’s the better fit.