An autonomous full-stack software creation engine that synthesizes, compiles, and hot-reloads React applications inside isolated E2B cloud microVMs through natural language dialogue. Features dynamic sub-1.2s Linux microVM sandbox provisioning, an infinite Vite supervisor daemon for zero-drop refactors, an autonomous function-calling LLM agent (file I/O, path normalization, shell execution), and an interactive split-pane studio with multi-device viewport toggles.

A multi-provider coding agent supporting Anthropic, OpenAI, and Gemini through a unified tool-calling interface. Includes a modular tool system for file editing, shell execution, search, and web access; semantic, episodic, and procedural memory for context recall across sessions; and parallel sub-agent orchestration for independent task delegation.

A full-stack perpetual futures exchange with an in-memory matching engine for order books, positions, and margin. Microservice communication is architected using Redis streams and pub/sub across the engine, API, and WebSocket layers, with live Binance price feeds driving funding rate calculations and liquidations. Features a Next.js trading dashboard with live charts, order book depth, and position tracking.

An atmospheric digital Punjabi trucker radio capturing late-night GT Road highway drives paired with classic Punjabi anthems. Features a headless YouTube audio engine wrapped in a vintage vinyl/cassette player with Punjabi truck-art aesthetics, real-time listener sync via Pusher presence channels, and interactive multi-sensory horn vibrations.

A real-time collaborative code editor with sub-100ms editing latency using Socket.IO rooms. Integrates the Monaco Editor to replicate a VS Code-like UX directly in the browser, with a backend architecture supporting shared sessions, live cursor sync, and scalable state sync logic for multi-user editing.

Hands-free browser automation built with the Web Speech API and Google Gemini. Enables natural language actions like navigation, search, and tab control through a modular system for voice intent parsing and action routing — reducing user interaction steps by up to 80% for common browsing tasks.