Plutos Equities is a data science project that predicts next-quarter financial metrics like EPS, net income, and gross profit for 80 NASDAQ companies. It combines structured financial data with unstructured text sources using BERT-based sentiment models, LSTM time-series forecasting, and gradient boosting. The system processes over 25 years of financial reports to deliver accurate, explainable, and data-driven predictions for corporate performance, built on one of the largest and cleanest financial datasets available in the market.
Relevant link: Plutos Equities Info

This research presents a comparative study on data privacy in Internet voting systems across Estonia, Switzerland, France, and Norway. It evaluates each system using five criteria—voter participation, computational cost, anonymity, availability, and transparency. The study assesses cryptographic techniques such as mixnets, homomorphic encryption, and zero-knowledge proofs, proposing a framework for minimum privacy and performance standards in future i-voting systems.
Research Paper: Data Privacy in Internet Voting Systems

This research, presented at MSR 2024, investigates how software developers integrate ChatGPT into their workflows. Analyzing 1,666 conversations from GitHub and HackerNews, we explored six research questions examining: common issue types, prompt patterns and their correlation with success, developer profile distributions, code modification practices, reproducibility of ChatGPT responses, and conversation structures. This work provides insights into AI-assisted software development and how conversational AI tools are reshaping the developer experience across diverse skill levels.
Challenge Link: 2024 MSR Challenge

CodeFume is an LLM-driven platform for detecting and refactoring object-oriented code smells through a two-stage prompt-based pipeline. The system identifies four prevalent smells (Long Method, Large Class, Feature Envy, and God Object) and generates automated refactorings grounded in SOLID principles. Evaluated on 100 Java and 54 C# samples, CodeFume achieves detection rates up to 100% and delivers refactorings with 94% functional equivalence, reducing cyclomatic complexity by up to 76%. The research also compares Google Gemini and GPT-4.1-nano, revealing trade-offs in maintainability preservation and token efficiency.
Research Paper: CodeFume: LLM-Powered Detection and Refactoring of Object-Oriented Code Smells

Developed a Agent Builder prototype enabling AI-powered form completion through a conversational interface. Designed and implemented the frontend for agent creation capturing parameters such as personality, behavior, and knowledge base. Engineered backend logic with Flask to process user inputs and generate contextual responses, implementing real-time messaging capabilities and robust error handling. Spearheaded development of a dynamic persona assignment system to enhance user engagement. Utilized Docker for containerization and conducted thorough testing to ensure system reliability.
GitHub Repository: Demo presentation

Falname is a Turkish coffee fortune-telling web application that combines image processing and LLMs. The system preprocesses user-uploaded coffee cup photos, detects objects through computer vision techniques, matches them with meanings from a traditional fortune-telling dictionary, and generates personalized fortune readings using language models. This project serves as a practical exploration of image processing pipelines while offering an engaging alternative to random fortune generators. Currently in development.

A comprehensive database-driven management platform for Sealanya Aquapark handling turnstile access control and customer ticketing operations. Built end-to-end to strengthen object-oriented thinking and full-stack development skills, the system features a fully designed MySQL relational database, backend APIs implemented with Python and Flask, and a clean HTML/CSS frontend served over localhost. I engineered the complete data model, implemented business logic for ticket validation and entry flow, and developed all UI and API layers, resulting in a fully functional, integrated operational system.

Hacettepe Lab Scheduler is a web-based reservation system designed to streamline equipment booking in university laboratories. The application provides an intuitive calendar interface that allows students and researchers to reserve lab equipment, view availability, and manage bookings efficiently. Built with a modern tech stack, the system aims to eliminate scheduling conflicts and improve resource utilization across Hacettepe University's laboratory facilities. Currently in development.

A collection of Python-based network simulation and protocol-handling scripts demonstrating core networking concepts. The projects cover socket programming, reliable data transfer mechanisms, congestion control algorithms, and packet routing implementations. Through modular, well-documented code, I gained hands-on experience with TCP/UDP semantics, packet loss and retransmission handling, and network layer abstractions, reinforcing theoretical coursework through practical applications.
GitHub Repository: Computer Networks Practice

A collection of machine learning tasks and experiments implementing classical ML algorithms and evaluation techniques using Python and Jupyter Notebooks. The projects cover data preprocessing, model training and evaluation, and comparative analysis of different ML models. Through hands-on implementation, I gained practical experience in data handling, feature engineering, model design and assessment, strengthening my understanding of end-to-end ML pipeline development and critical evaluation of model performance.
GitHub Repository: Machine Learning Practice

This project addresses transportation challenges on campus, where inconvenient shuttle schedules and insufficient services are common. Our goal is to create a safe and eco-friendly solution that connects students with personal vehicle owners heading in the same direction. By facilitating these connections, I aim to reduce the number of cars on the road while providing reliable rides for students. The app allows hitchhikers to post ride requests, which drivers can accept, with all interactions recorded for safety. This approach not only enhances transportation options but also fosters social interaction within the campus community.

This project aims to create a secondhand sales platform primarily for students, allowing them to buy and sell items among each other. Users can create posts for items they wish to sell, and other users can comment or send purchase offers on these listings. The platform also includes a lost and found feature, helping users reconnect with lost belongings. A real-time chat feature enhances communication, making negotiations easy and efficient. The responsive design ensures a user-friendly experience across all devices, while prioritizing data privacy to protect users’ personal information.
Relevant link: https://bilco.web.app/

This project aims to create a comprehensive ticket sales platform with multiple user types. Regular users are the target audience who purchase tickets, while organizer users create and sell event tickets. Worker bee users address issues reported by other user types and the admin user is responsible for generating daily reports, managing embedded system data. The application involves numerous transformations and complex inter-object relationships, with a strong emphasis on database design and management to support efficient operations and user interactions.
GitHub Repository: https://github.com/.../ticketify

This project, Pine, is a fully functional programming language with distinctive features. Its syntax is defined using BNF notation, which provides a clear and precise way to describe the language's structure. To enhance its functionality, I utilized Lex for tokenization and Yacc for parsing, ensuring a robust and efficient language framework. The combination of these tools allows for effective error handling and enhances the overall performance of Pine, making it a powerful tool for developers.
GitHub Repository: https://github.com/.../Pine

These projects showcase my ability to handle complex, low-level computing tasks and digital design. MIPS Assembly Labs cover core MIPS concepts like control flow, recursion, and I/O. Digital Design Labs focus on Verilog, exploring logic gates, ALUs, and memory design, with practical simulations of both combinational and sequential logic. The Operating System Project involves systems programming in C, covering system calls, multithreading using POSIX threads, and TCP socket programming.
GitHub Repository: MIPS-Assembly_labs
GitHub Repository: Digital-Design
GitHub Repository: Operating-System

This section showcases a personal archive of small projects developed using object-oriented programming (OOP) principles in Java and C++. These projects serve as practical examples of coding practices, illustrating the application of OOP concepts in real-world scenarios.

Java GitHub Repository: java_algo_problems
Java GitHub Repository: java_game_1
C++ GitHub Repository: cpp_cs_fundamentals
C++ GitHub Repository: cpp_algo_problems