Luka Onikadze

Colorize is an open-source tool that brings color to black-and-white images using the DeOldify model. Initially, I was hired as a freelancer by a public archive to develop this tool for colorizing historical photographs. After completing the project, I received permission to release it as open-source, making it available to the community.

The project features a front end (FE) for image uploads and a back end (BE) for the colorization process. With straightforward setup instructions, you can easily deploy the tool and start exploring the intersection of history and technology.

Tensoreye is a web-based application that enables users to construct neural networks through an intuitive graphical user interface (GUI). Leveraging TensorFlow.js for its back end, Tensoreye facilitates the creation and deployment of neural network models. I developed this application entirely, ensuring a seamless and user-friendly experience. The primary purpose of this project was to educate students and provide a platform for practice, offering a hands-on approach to learning and experimenting with neural networks.

A monolithic application, written on C# and AngularJS, which had tremendous performance and scalability issues. The management team had zero confidence that it could launch it in one year. I took over the project one year before its release.

During that period, my team and I rewrote and decoupled the monolithic architecture into the service-oriented one. We added a Node.js web server between the front end and the microservices, integrated technologies like Elasticsearch, Redis, and Elastic for caching. We dropped the whole project from AngularJS and rewrote it on Angular because AngularJS lacked documentation and stability. This company is now in the top five in Georgia's sampling industry.

This app required a lot of lazy loading to meet its business demands for dynamic rendering compression and web caching. It had to load tons of data and renew itself every few seconds.

I wrote this myself, using Angular lazy loading for dynamic routing. I also wrote a caching mechanism using IndexDB, integrated a service worker for PWA, and used this technology for asset caching. Additionally, I created a dynamic rendering mechanism, which rendered chunks of data without static height, integrated Socket.IO, and made a request-response pattern using observable streams.

A fast betting app that my team and I built on PixiJS. It was a whole new world for me in web development because I had no prior experience in canvas-based applications. After studying the canvas world, I decided to use the PixiJS framework to deliver business demands.

The lobby itself is written on Angular, and the rendering engine is on PixiJS. After releasing a few games like this, we created a dedicated library on PixiJS. The library was generally purposed and had all the functionalities for rendering and adding assets and managing game logic for any new betting application.

Built an enterprise-level eCommerce application in Angular from scratch. Used all modern tools, including an Nx workspace monorepo pattern, to split and manage the huge codebase.
I led the front-end team and worked with PDMs to plan and execute the sprint goals in the most performant way. I made a critical decision, which was not evident then, and gained us a huge boost and benefit later in the project.
I also kept the highest quality of code as much as possible. I was a gatekeeper in merging the code and reviewing it extensively.

Led the team from moving a web SPA to the native app, using modern tools like Ionic with Capacitor. Used a monorepo pattern (Nx workspace) to avoid code duplication during the development and for further use.

I developed a web-based translator app that provides seamless translations between Georgian and English. Leveraging LangChain and OpenAI's large language model (LLM), the app delivers highly accurate and context-aware translations in real time. The user-friendly interface allows for easy navigation, making it accessible for both casual users and professionals. This project showcases my ability to integrate cutting-edge AI technologies into practical applications, providing a reliable tool for communication across these two languages.