Technology should empower people, not gatekeep them behind expensive subscriptions or complex coding requirements.
For the past decade, the industry has pushed us toward renting software from massive cloud companies. This has resulted in slow apps, compromised privacy, and rising costs for everyday users and startups.
My mission is to reverse this trend. I build community-driven, local-first software. These are tools that run incredibly fast on the computer you already own, respect your privacy by never sending your data to the cloud, and are built to be accessible for everyone.
This vision is actively being built through three core pillars of development, designed to put power back into the hands of the community.
Instead of relying on distant servers to process your video and audio, I build applications that tap directly into your local hardware (like your graphics card). This means zero lag, no monthly API bills, and the absolute guarantee that your data never leaves your room.
Live Example: RenderCaption
RenderCaption is a fully deployed, free desktop app that instantly generates audio transcriptions and subtitles. It features a beautiful, zero-code graphical interface so anyone can use state-of-the-art AI without needing to write a single line of code.
In Development: Whisper Studio
An upcoming community-focused video editing suite. It builds upon RenderCaption’s core technology to provide professional-grade video tools that operate entirely offline, democratizing high-end video production.
Intelligence should not be outsourced to trillion-dollar companies. By utilizing Small Language Models (SLMs), we can run powerful AI directly on our own devices or at the network edge.
In Development: Nullshore
An exploration into building a new kind of search engine for the open-source community. Nullshore uses AI to help developers and founders instantly discover obscure, highly complex deep-tech infrastructure, accelerating global innovation by making open-source knowledge truly accessible.
When we do build for the web, it must be fast and modular. I work with brands and agencies to replace their fragile legacy systems (like massive, broken Excel sheets) with modern, blazing-fast web architectures.
Live Implementations
By distributing logic globally using Next.js and Cloudflare, I help businesses scale efficiently. Every project is built with a signature modern, dark-mode aesthetic to prove that powerful utility software can also be beautiful.
To ensure developers and users around the world have access to top-tier AI tools, I actively convert and optimize massive, state-of-the-art speech models into the highly efficient .gguf format. This allows anyone to run them locally, for free.
- Parakeet-TDT 0.6B Multilingual GGUF
(Supports 25+ European languages)
- IndicConformer GGUF
(Optimized for Hindi and Punjabi)
- Portuguese FastConformer Hybrid GGUF
(Optimized for Portuguese ASR)
- Hinglish Conformer CTC GGUF
(Optimized for Hindi-English code-switching)
For those interested in the technical stack powering this ecosystem:
Core Stack: Rust | Tauri | Next.js 15 | Cloudflare Workers | Supabase
AI Systems: C++ | Whisper.cpp | GGUF Quantization | ONNX Runtime | CUDA
Search Indexing: Quantized GGUF binaries, Offline speech-to-text models, Local privacy-first audio transcription, Parakeet-CPP, NVIDIA NeMo ASR conversion, FastConformer Hybrid, Conformer CTC, Vulkan accelerated inference, Whisper alternative.