Project 01
Smart Desktop Automation
Show a repetitive desktop task once. The agent learns it, takes over the computer and keeps going on its own until the task is done, even when something unexpected goes wrong.
Role: Product design, Full stack development, AI integration
Stack: Python, Playwright, OpenCV, LLM
Overview
Teams logged into the same portals every day and downloaded the same reports by hand. When a page didn't load or a window didn't open, the work silently stopped halfway.
Goal
Build an agent that learns a task from a single demonstration, catches and resolves errors on its own and runs smoothly on both Windows and macOS.
Task teaching
The user demonstrates the job once; the system records clicks, typing and waiting points and turns them into a repeatable task.
Error recovery
It catches cases like a page not loading, a window not opening or an expected button missing, and applies recovery steps such as reloading, retrying or taking an alternative path.
Cross-platform
The same task runs on Windows and macOS, on Intel and Apple Silicon (M series) processors, without changes.