I built a fully local AI assistant at 16 — no cloud, no API keys, runs on your GPU

Short summary

A 16-year-old from Pune built O-AI, a fully local AI desktop assistant running entirely on-device via llama.cpp and Ollama with zero internet dependency. Features voice control in English, Hindi, and Marathi, 30+ desktop automation paths, and a self-learning core that fine-tunes locally. Open-source with technical lessons on validation, GPU optimization, and multi-step agent design.

• •Fully local AI assistant running on GPU with no cloud or API keys required
• •Supports voice control in three languages with desktop automation and screen vision capabilities
• •Open-source implementation with detailed technical lessons on validation and agent design

Generated with AI, which can make mistakes.