Avi Aryan

A Python-based, real-time dictation tool for M-series MacBooks. It uses the open-source Parakeet voice-recognition libraries from Nvidia for on-device voice transcription. Then we use local Gemma 3 to clean up the text, if configured. The result is state-of-the-art voice dictation that is more accurate than what comes built-in on Apple devices. Open-source with over 60 stars on GitHub.

Python-based Telegram bot that transcribes and summarizes voice notes using the latest AI models. It uses Groq-hosted Whisper to transcribe voice notes, then summarizes the transcription with Llama 3.3, also running on Groq. This helps us deliver high-performance audio summarization by using open-source models running on one of the fastest AI inference platforms, Groq. This open-source project has over 70 stars on GitHub.

Clipjump is a multiple-clipboard management software for Windows. It was built to make working with multiple clipboards super fast and easy. The program records changes in the system clipboard, stores them without limits, and provides innovative ways to work with them. The project, which I built solely, has garnered over 100,000 downloads to date, with more than 50 public releases.

An Electron-based app that allows you to speak using your voice. It will transcribe what you say in real-time with high accuracy using Whisper.cpp's streaming module. Moreover, it uses Llama 3 via Groq to fix punctuation, capitalization, and grammatical issues in the transcription. In the end, you get a perfect transcription of your voice in real time.

BigEval is an npm library that is an alternative to JavaScript's eval() for solving mathematical expressions. It can be extended to use the Big Number libraries available to provide results with maximum precision.

AnyAudio is an open source music streaming platform. It uses the massive dataset of YouTube to allow you to listen to any song you want. For downloading data from YouTube, it uses youtube-dl and to serve the data to the end users, it has a beautifully designed React progressive web app.

VSCode Notebook is an attempt to use VS Code as a complete note-taking application. It includes several cool features such as notes encryption, fast search, hierarchical organization, and an auto Git backup facility.

I created a Telegram bot that saves user forwards and other messages to their Readwise account. It is a B2C subscription-based service. We store the user's sensitive API key data in an encrypted-at-rest Supabase database.

GoRemote is a remote jobs aggregator for the Indian time zone. We parse job listings on Stack Overflow and other sites, detect any time zone requirements in that job, and list only those jobs where Indians can apply.