Gokhan Yaliniz

I designed a framework similar to the NestJS framework that served as a project software structure generation tool for modular vision systems. It includes a CLI for generating controllers, services, adaptors, UIs, and test software.

The framework helped generate new application software with already implemented features and different configurations in seconds, and the scalability of the software projects increased substantially.

Eliminated developing custom software for projects repetitively, which helped save engineering time.

I designed a no-code web app as a GUI generation tool. Our users could generate custom GUIs at will, and their structure was standardized, which helped users adapt to them quickly, and developers saved time by avoiding building custom GUIs.

I designed the latest architecture for embedded systems that is scalable, highly reusable, and maintainable.

I was the system's main developer, implementing the core application software and leading a team of six developers to scale the system for all the product families in my department.

My team saved 80% of engineering time with the architecture I designed for more creative tasks. We created an API, UI, testing, and adaptor software for systems, which made up 60% of the codebase and was automated.

The architecture led to the automation of creating GUIs for projects, and it also helped to standardize communication protocols and code structure. It also helped to generate software systems by using the source codes of projects as configuration files.

I developed a console application that analyzed objects and their relations in software to structure UML diagrams of projects.

By generating UML diagrams of apps via this app, the software's architecture was better grasped by all stakeholders, less experienced developers were integrated into projects faster, diagrams of projects were standardized, and documentation processes took less time.

I designed and built an unsupervised video summarization tool for my MSc thesis using independently RNNs for a reinforcement learning-based method. I implemented the algorithm in Python and used the PyTorch framework.

Based on the experiments conducted on two benchmark datasets, we observed that better summarization performance could be obtained compared to the state-of-the-art methods on a video summarization task.

The project was partially supported by a grant from the TUBITAK 1003 Primary Subjects R&D Funding Program.