Open-source contribution

SapinSapin AI

Volunteer work on an open Philippine-language AI initiative: parts of its public site and speech demo, and Sappy, the assistant I created to answer questions about the project in people's own language.

Visit the site (opens in a new tab)See the code (opens in a new tab)
Built for
SapinSapin AI, an open-source community initiative for Philippine-language AI, founded by Tim Santos
My roles
  • Creator & AI Systems EngineerSappy
  • Designer & Front-End DeveloperSapinSapin AI public site
When
August 2026 – Present

The situation

SapinSapin AI publishes its datasets and models in a technical catalogue that mostly researchers can read. Newcomers needed a way in that a curious visitor could follow, and as the community grew, its volunteers kept answering the same questions by hand, in Discord and elsewhere.

What I built

I started on the public site: designing and building parts of it, including a demo that recognises speech and reads text aloud, straight from the browser, with every catalogue figure re-pulled from the live source on each rebuild. Then I created Sappy. When someone asks a question, Sappy first looks up the relevant passages in a knowledge base I prepared and keep up to date — one that separates confirmed facts from experiments and claims nobody has checked yet — and only then writes an answer from them. The same Sappy answers on the website and in the project's Discord, where I also designed how newcomers join: introduce yourself, get the Member role, and the rest of the server opens up.

An unpaid, open-source contribution. I designed and developed parts of SapinSapin AI's public web experience and browser demo, then created Sappy, which I built and maintain as its AI systems engineer. The initiative itself was founded and is led by Tim Santos; Sappy is my part of it.

What it lets people do

  1. 01Speak into the page and see it transcribed, or hear text read aloud, in ten languages spoken in the Philippines
  2. 02Ask Sappy about the project on the website or in Discord, in your own language, and see how many passages its answer drew on
  3. 03Join the Discord community through an introduction step that opens up full participation
  4. 04Read catalogue figures that are re-pulled from the live source on every deploy

Featured in

NYONYOMarc is also a contributor to NYO
Testing and feedback, internal code, and UI/UX work on the LLM router this runs on. (opens in a new tab)

Under the hood

The tools this is built with. You do not need to know any of them to use it — open one if you are curious what it does.

How a project goes

  1. We talk about the problem

    Not the software — the problem. What is slow, what gets forgotten, what people keep having to redo by hand. Usually one call is enough to see the shape of it.

  2. I write down what I heard

    In plain language, before any building starts, so you can correct me while it is still cheap to correct. You approve it or you don't.

  3. You see it early and often

    Working pieces, not screenshots or promises. If something is wrong, you say so while there is still time to change it.

  4. It goes live and stays supported

    Hosting, domain and setup handled. The Valley O'Ville site has been running since 2023 — I don't disappear after launch.

Have something in mind?

Tell me what you’re working through.

Not ready to talk yet? Read the newsletter, a monthly letter on what I am building and learning.