What we're building
Built by Sikh builders. For the Panth.
Parmaan Khoji (Gurbani Semantic Search) is an AI-powered search engine designed to find thematically deeper connections—known as *parmaans*—across Sikh scriptures, including Sri Guru Granth Sahib Ji and the Dasam Granth. Instead of relying on rigid keyword matching, it uses advanced contextual stanza embeddings to understand the core theological meaning of a query within the context of its entire stanza. Users can seamlessly search using Punjabi, English, or Faridkot Teeka translations and receive highly accurate results thanks to a two-stage AI pipeline featuring cross-encoder reranking. Furthermore, the system continuously improves through real-time user feedback and offers instant, cached results without waiting for model inference.
A searchable Punjabi-English dictionary focused on Gurbani terminology, Sikh historical vocabulary, and common Gurmukhi words — built to help the next generation connect with the language of their heritage.
A experimental sandbox for testing new features, integrations, and UI components before they ship to production on the Sikh Tech Collective platform.
Audio Splitter — Built for Kirtan Seva Managing kirtan recordings is tedious. After every program, someone has to sit down with a full 2-3 hour audio file, manually cut it into individual tracks, rename each one, and share it out — all before the recording even reaches the kirtanis. This tool was built to make that entire process take minutes instead of hours. Audio Splitter is a local web app that lets you upload any audio file (MP3, WAV, MP4, M4A), define timestamps for each kirtani, and instantly download a clean ZIP of individually named tracks. It supports YouTube and SoundCloud URL imports, bulk timestamp pasting from your notes app, and smart autocomplete for Bhai/Bibi naming conventions — including your project name automatically appended to every track title. For those already in the Google Drive or SoundCloud ecosystem, there are direct upload integrations so tracks land exactly where your team needs them. Built with Flask and ffmpeg under the hood, splits are near-instant — no re-encoding, just fast byte-level extraction. Designed to run locally on any Mac with Python and ffmpeg installed.
