Mark Kiraly

Project overview: An Rnote fork adding LaTeX export to handwritten note-taking.

Technologies: Rust, GTK4, pdflatex, Google AI Studio

• Project description: An enhanced fork of the open-source Rnote sketching app that introduces an experimental pipeline to convert handwritten notes directly into structured, compiled LaTeX documents. Further development ideas remain as to-dos.
  Source code(s):  Package link:

Project overview: A simple Android app that provides background hints for math problems. It supports creative thinking without revealing key ideas or full solutions.

Technologies: Android Java, LaTeX rendering, Google AI Studio

• Project description: The app stores KöMaL problems (https://www.komal.hu/home.e.shtml) in plain text and shows them as LaTeX text. Users can pick problems at random, and hints are given on request. Hints only provide general background knowledge and the necessary tools, without offering ideas, strategies. The app and its database are highly scalable.
Source code(s):

1. Subproject overview: muzplayer - Tiny music player with custom randomization streaming algorithm

Technologies: C++ (Qt6 framework), CMake

• Subproject description: A simple, Qt-based lightweight music player supporting ogg, mp3, wav, flac, m4a, and aac formats. Randomly selects files from the chosen folder and avoids repeating tracks played within the last 10 minutes.
  Source code(s):  Package link:

2. Subproject overview: lingua - Speech translator primarily for language learning. Languages: English ⇋ Hungarian

Technologies: Python (PyQt6), openai-whisper

• Subproject description: Real-time speech-to-text and translation application using the OpenAI Whisper model. Transcribes and translates English ⇋ Hungarian conversations in near real time, suitable for language learning.
  Source code(s):  Package link:

Project overview: Simple Linux desktop app with GUI for searching public docs archives.

Technologies: JavaScript (Electron), HTML file processing

• Project description: This application is a simple and fast desktop tool to search, download, and manage books from distributed public databases. Features include multi-language support (English and Hungarian), easy navigation, download management, and a built-in search engine. Before downloading, the application searches for the server with the best ping.
  Source code(s):  Package link:

Project overview: Theory of Computation project in progress, using a JavaFX GUI.

Technologies: Java (JavaFX, Maven)

• Project description: This project is based on the topics of FSMs (Finite State Machines) and regular expressions. The user can graphically edit an FSM (NFA or DFA) for a given regex pattern (or sample strings), and then verify its correctness. But the program can generate regex pattern with sample strings.
  Source code(s):

Project overview: Travel expense form software designed to reduce administrative workload.

Technologies: Java (Spring framework), JasperReport, JavaScript (BackboneJS), Postgres, H2DB

• Project description: This software allows users to quickly and electronically complete travel expense forms by automating repetitive manual tasks. Users can fill out the forms through a web interface and then download and print the results as PDF.
  Source code(s):

Project overview: Solved advanced programming problems published in the KöMaL journal.

Technologies: Python, C++, Haskell, C++ with CUDA

• Project description:
  
  Problem S.20 (Python):
  This was my first meaningful GitHub project. The task was to find the smallest possible circle covering n given points in the plane. The tools for solving the problem are primitive, but because of that, the approach is complex.
  Source code(s):
  
  Problem S.141 (C++ / Haskell / C++ with CUDA):
  I create a simple mathematical model of the problem and parallelize it. It is essential to emphasize that our goal is not merely calculation but rather the rapid computation, especially for large numbers.
  Source code(s):
  
  Based on Problem S.127 (C++):
  This snippets is a generalization of task KöMaL S.127.
  Source code(s):

1. Subproject overview: Experimental Refinements: State-Space Model Architectures

• Subproject description: Investigating the capabilities of State-Space Models as efficient alternatives to Transformers. I am experimenting with architectural modifications to improve their performance and stability.
  Source code(s):

2. Subproject overview: Industrial Anomaly Detection: Robust & Training-Free Inspection

• Subproject description: Developing a highly robust, training-free method for industrial visual inspection by refining memory-bank based algorithms, ensuring the resulting system adapts seamlessly to new products without optimization loops.
  Source code(s):

Technologies: Java (Swing), XML, Google Cloud Platform

1. Subproject overview: Economical Desktop App

• Subroject description: This application stores the specified expenses and income in an XML file or, in the case of an internet connection, on Google Drive. In its absence, it stores the data on the computer. Generates statistics that can be exported as Excel (.xlsx) files.
  Source code(s):
  

2. Subproject overview: Language Dictionary Desktop App

• Subproject description: This application provides a graphical interface to view and manage an XML-based dictionary.
  Source code(s):
  

3. Subproject overview: Simple Chat App

• Subproject description: The application allows users to exchange messages in near real-time with Google Drive API, although this approach is slower than socket-based communication.
  Source code(s):

1. Subproject overview: Dataracing Challenges

Technologies: Python (NumPy, Pandas, scikit-learn, Optuna)

• Subproject description: Developed solutions for the 2022 and 2023 dataracing competitions using various data science methods and optimization techniques.
  Source code(s):
  

2. Subproject overview: KutyAI - Dog Bread Recognizer

Technologies: Python (Tensorflow, PyQt)

• Subproject description: This is a Python graphical application for dog breed recognization, featuring 420 breeds and 42,000 images. For some reason, the model’s accuracy cannot exceed a certain percentage with this dataset, likely due to noisy images automatically scraped from the web — for example, 'Bolognese' could refer to both a dog breed and a dish. Nevertheless, the results are interpretable: purebred dogs are recognized very well, e.g., a Labrador is correctly identified as a Labrador about 90% of the time.
  Source code(s):

Project overview: N/A

Technologies: Java (Spring framework), Postgres, JavaScript (VueJS, React, Angular)

• Project description: This repository contains multiple frontends built using various frameworks, all connecting to the same backend. The backend is a system module that handles transactions. I use this repository to showcase my basic frontend skills. It has helped me a lot in understanding the logic behind the most popular frontend frameworks.
  Source code(s):

Project overview: N/A

Technologies: HTML, Bootstrap CSS, native JavaScript

• Project description: Self-designed responsive website. Serves as an earlier portfolio, though its design now feels slightly outdated.
  Source code(s):

1. Subproject overview: Numerical algorithms in MATLAB

Technologies: MATLAB

• Project description: Implemented various numerical algorithms in MATLAB, including affine transforms, computer arithmetic, Gauss elimination, interpolation, iterative methods for solving linear and nonlinear systems, numerical integration, and QR decomposition.
  Source code(s):

2. Subproject overview: Visualization in VisIt software

Technologies: Python (NumPy, PyVista), VTK format, VisIt

• Project description: This repository demonstrates how to compute the arc length of a function curve using VTK files, in cases where symbolic (often irrational) integrals are difficult or impossible to calculate.
  Source code(s):

Project overview: This project analyzes North Korea's geographic infrastructure and missile test data to visualize spatial relationships and launch statistics.

Technologies: Python (Pandas, GeoPandas, SQLAlchemy, etc.), Postgres, QGIS, PostGIS, shapefiles,

• Project description: This repository contains materials for a data analysis project focusing on North Korea's geographic and missile test data, created as part of a university course (PTE TTK Data Analysis). The project demonstrates skills in GIS analysis, PostgreSQL/PostGIS, Python data processing, and visualization using QGIS and Pandas.
  Source code(s):

Project overview: A video game mod development project.

Technologies: Java (Gradle), Fabric API, Blockbench

• Project description: Developed a Minecraft Fabric mod that visualizes sorting algorithms step by step using chests as arrays and item stacks as values, and added a new custom mob, whose appearance I designed in Blockbench.
  Source code(s):