Work
Systems built for other people, and still running. Each one leads with the decision that mattered rather than a list of features.
Matching a hundred students to fifty mentors, live, on venue wi-fi
One evening, fifty mentors, a hundred students, and no good way to decide who should talk to whom. Students now answer a short questionnaire and get a ranked shortlist with a reason for each match — and the whole thing runs with no AI calls, no accounts and no login.
Read the case study
Every public high school in Kansas, and who to call at each one
They booked speaking engagements one school at a time, by searching for a phone number. The directory now holds 328 bookable venues with a named contact for each, rebuilt monthly from public records — and it refuses to collect anything a school chose not to publish.
Read the case study
Turning public filings into a queue somebody can actually work
The information was all public and none of it was usable — scattered across federal filing systems in a dozen formats, none of which agree on how to spell a person's name. It now arrives as a deduplicated, scored queue where every entry carries a link back to the filing it came from.
Read the case study
A nonprofit whose database was a spreadsheet
Applicants and cohorts tracked in workbooks that drifted out of sync, and nowhere to send anyone who asked about the programme. Now an application system, a student portal, and an admin console the staff run themselves.
Read the case study
