BuildFest
1st place of 44 teams
EveryMind's Choice award
A one-week challenge to turn EveryMind's 988 training feedback into a working tool that shows what is landing with participants and where the gaps are. Our team's tool took first place of 44 teams and won the EveryMind's Choice award.
EveryMind is a Maryland mental health nonprofit that runs trainings across the state on 988, the national suicide and crisis lifeline. Everyone who attends fills out a survey afterward. Our tool reads those survey exports and answers four questions: did the training work, for whom, where are the gaps, and which sessions need a look. One command, or one drag-and-drop in the browser, turns an export into a single dashboard file that opens offline without a server, so EveryMind's staff can use it without installing anything.
- 200
- students competing
- 1,272
- survey responses analyzed in about 10 seconds




The two certificates, presented by the Smith School's Center for Social Value Creation, Hack4Impact UMD and the Claude Builder Club at Maryland.
Select a step.
After each training, participants answer a Microsoft Forms survey: what they knew about 988 before and after, how confident they feel using it, and open-ended comments. EveryMind exports the responses to Excel. The analysis cleans the export, removes repeats, merges survey versions and measures how much knowledge changed, by group and by session. A new month, county or survey version needs no code changes. Everything lands in one HTML file that opens in any browser without an internet connection. A new export can be dragged straight onto the page, and it is read in the browser and never uploaded. From the dashboard: a one-page summary for funders and the board, session report cards, advice for the next version of the survey, and pre-filled survey links and QR codes for each session.The tool
Screenshots from the team's public demo at everymind.benmross.com. It turns on its own, and you can drag or swipe it sideways or use the arrows, or hover over it to hold it still.






What I built
- EveryMind redesign
- I rebuilt the dashboard in EveryMind's own look: their colors, their fonts (embedded so it still works offline) and their website's layout and navigation, so it reads as their tool rather than a student project.
- Funder and board summary
- A printable one-page summary with a July to June fiscal-year selector and a grant-report table. Groups under 11 people are combined and no comments are shown.
- Role views and a trainer playbook
- "Viewing as" switches the dashboard between the people who use it, and trainers get a playbook on the Sessions tab.
- Session links
- A builder that makes a pre-filled survey link and a QR code for each training session, all generated offline.
- Answers for participants
- Draft answers to the eight questions participants ask most, for trainers and a participant FAQ, written for EveryMind to approve.
- Honest labels
- A 988 help footer on every page, and open-text results reported as agreement with AI-drafted labels, not as accuracy.
The rest of the build
My teammates built the Python analysis (cleaning, paired statistics, group tables and session scorecards), an in-browser engine that matches it check for check, and pages for participants and staff.
Privacy
The survey answers are real feedback from real people, so the dashboard embeds no names, comments, IDs or timestamps, and any group under 30 people is marked too small to compare. The screenshots on this page come from the team's public demo, which shows only those summaries.
How it was judged
Industry judges in AI and data, with a representative from EveryMind, scored business and social impact (35%), technical soundness (30%), functionality (20%) and AI process and transparency (15%). The top five teams presented at the finals.