Machina Mundi is an organization working across Greater Hartford, Connecticut, in Hartford, East Hartford, Manchester, and South Windsor. We run hands-on artificial intelligence programs for underrepresented students, and we developed ConvoAAVE, a speech corpus of African American Vernacular English built to improve how speech recognition systems handle the dialect. That work was named a regional finalist for the Princeton Prize in Race Relations in May 2026.
Machina Mundi began with a straightforward observation. As artificial intelligence became central to education, employment, and daily life, students in Hartford and East Hartford had far less access to the tools and instruction available in better resourced districts nearby. The gap was one of access, not of interest or ability.
We began with Code Ninjas, and later with YMCA programs and host schools, building an introductory summer course delivered in three three-hour sessions. The aim was not only to teach coding, but to show students how these systems work, how they affect daily life, and how students might build them.
By the end of the first round, nearly 900 students had trained simple models, examined how chatbots generate responses, and discussed how AI systems reach decisions. As the sessions continued, students asked why voice assistants and automatic captions so often misread the way they spoke, while handling other speakers without difficulty.
That question pointed to a second problem. Speech recognition systems learn from recorded audio, and the audio used to train them underrepresents many Black communities, including speakers of African American Vernacular English. The result is a bias of representation rather than intent: absent from the training data, a community is misheard.
That finding shaped the organization’s current direction. Alongside teaching students about AI, we now work to improve it, and that work became ConvoAAVE, our African American Vernacular English speech corpus.
The team presented ConvoAAVE at the Princeton Prize in Race Relations ceremony in May 2026, where the organization was named a regional finalist.
Community partners:
"If AI is going to serve everyone, it has to be trained to understand everyone. Language is part of identity, culture, and community, and that has to show up in the data."
Seven places to look, depending on whether you came for the research, the classroom programs, or the policy. The last one is a room you can walk around in.
Speech corpus
More than 80,000 words of African American Vernacular English speech transcript, built to improve how speech recognition handles the dialect. Recordings stay private; the transcripts are published.
Read the methodology 02 EducationHands-on sessions with Code Ninjas, the YMCA, and host schools. Nearly 900 students since 2024.
See the programs 03 Federal workA $5,000 per-school AI Equity Grant provision, drafted with the office of Congressman John Larson.
Read the provision 04 2024 to 2028Where the organization has been, from the first Hartford sessions to deployment support ahead.
Follow the arc 05 The teamThe students who run the sessions and collect the data, and the advisers behind the curriculum.
Meet the team 06 UpdatesNotes from the team, month by month, including the May 2026 meeting with Congressman Larson.
Read the latestA second site, built as a place rather than a page: seven chapters you move through with a drag or the arrow keys, objects you can pick up and read, and a guestbook on a floating scroll where visitors leave a message of their own.
Enter the RoomClick and hold the doors
Machina Mundi runs on volunteers and partner classrooms across Greater Hartford.
