We're a nonprofit collective of high schoolers and undergrads working on research, software, and community projects across machine learning, language preservation, education, and accessibility — and giving everything we make back to the public.
One Sphere Array Inc. is run by students, for the public. We don't have a single product or vertical — we have a worldview: that research, education, and software should be made by the people they serve, and shared without gatekeeping. We work across a range of ventures depending on what the moment calls for.
Code, datasets, papers, and process notes go out under permissive licenses. If we make it, you can use it.
We pick problems because they matter to a community we know — not because they're trending on a leaderboard.
Everything ships from young people learning out loud. Mistakes included, written down, and learned from publicly.
Our latest set of long-running projects designed by students and shipped openly. Most are still works in progress.
Datasets, embeddings, and small language models for Dzongkha and other under-served languages of the Himalayas — built with native speakers, not despite them.
Workshops, reading groups, and original research papers aimed at students who don't have a research lab next door. Free, asynchronous, and rigorous.
Small, free utilities — assistive tools, study aids, language helpers — built for users who get ignored by the rest of the software industry.
An experiment in using LLMs to build cross-language knowledge graphs that help under-resourced communities surface and connect their own information.
We're deliberately not a single-vertical organization. Different ventures call for different muscles, and we like it that way.
Small models, applied research, and honest evaluation — over hype.
Especially for languages and dialects that don't get attention.
Free curricula, peer mentorship, and pathways into research.
Archives, oral histories, digitization, preservation tools.
Useful, accessible, free. Minimum viable, maximum reach.
Designing first for users that mainstream tooling ignores.
Original investigations, written up and shared in full.
Mentorship, hosting, programs, and just being good neighbors.
we're not a startup or a data company.
we're a group of young people who think research, software, and learning should belong
to the public, not large corporations.
We're a very small team and we read everything. Reach out about partnerships, questions, joining a venture, or just to say hi.