Machine Learning engineer Artem Orlovsky has developed a personal AI that can autonomously create web applications, manage their functionality, and aggregate large data sets. Initially, the system was created for learning foreign languages, but it quickly evolved into a full-fledged AI assistant that writes code itself.
Machine Learning engineer Artem Orlovsky has developed a personal AI that can autonomously create web applications, manage their functionality, and aggregate large data sets. Initially, the system was created for learning foreign languages, but it quickly evolved into a full-fledged AI assistant that writes code itself.
In a comment to dev.ua, the developer said that the idea for the project arose due to the shortcomings of existing solutions on the market. According to Artem, he decided to write his own agent as a simpler analogue of the well-known OpenClaw system, correcting the technical aspects to suit his needs. The primary task of the AI was to help the engineer learn Spanish via messenger, while providing a personalized approach with voice messages and exercises.
Later, the developer expanded the bot's capabilities and instructed it to aggregate news in the form of a daily digest. When the information in the messenger became too much, Artem asked the agent to develop a separate website for convenient reading, and the AI independently wrote the code and created the interface.
"Over time, new ideas appear, and when I describe these features to the agent in the messenger, he actively creates new convenient functionality on the website and manages the relevance of the information," the engineer notes.
Recently, the system has received another practical application - an aggregator of opportunities for Ukrainian scientists, which searches for relevant hackathons, scholarships and grants. The developer was inspired by this idea after a conversation with the co-founder of the AI platform Lapathoniia , Bohdan Didenko, about finding funding and computing power necessary for training local language models, in particular Lapa LM. According to Artem, after realizing the need, the algorithm integrated this complex functionality into the site in just one day.
The project is currently at the Proof of Concept stage and operates as a personalized tool physically hosted on a Raspberry Pi microcomputer. In addition to the main functionality, the system already includes technical dashboards for monitoring the status of the site. However, the author is considering a public release in the future. For full deployment and access to all interested parties, the team will need to think through the architecture and move to separate servers.
