Muhammed İkbal KARADELİ

• I work as a data scientist and AI Specialist to develop AI-powered technologies to improve cybersecurity solutions.
• A project was developed to provide proactive protection against cyberattacks, designing a comprehensive solution that thoroughly examines user behavior to assess potential threats.
• In this context, 26 'detection' modules were created using machine learning approaches-particularly unsupervised learning.
• All detection mechanisms employ an algorithm and scoring system based on behavioral anomaly analysis; by evaluating these scores, the potential risk posed by a specific user or IP address is reliably determined.
• The system was developed using Python to connect to PostgreSQL and Elasticsearch databases, supporting data insertion, updating, deletion, and the creation of new databases.
• On the Java side, a similar application was implemented with Spring Boot Rest API.
• All these components were containerized with Docker and successfully deployed to an Azure environment.
• During the project, extensive expertise was gained in system-level technologies such as Docker, Elasticsearch, and PostgreSQL.
• Additionally, significant proficiency was achieved in designing detection algorithms and implementing unsupervised learning models.
• All these efforts were carried out without any external consultancy, from initial concept to final deployment.
• Gained hands-on expertise in cloud-based systems, including their architecture and operational workflows.
• Key words: Data Science, Artificial Intelligence (AI), Cybersecurity Analytics, Behavioral Anomaly Detection, Unsupervised Machine Learning, Detection Algorithm Development, PostgreSQL, Elasticsearch, Spring Boot (Java), RESTful API Development, Docker & Containerization, Microsoft Azure Deployment, Cloud Computing & Architecture, End-to-End System Development, Python Programming.

• I worked in the academic research part of the research project established by Dr. Zehra Gülru ÇAM TAŞKIRAN for the following purposes:
• An advanced hyperchaotic image encryption algorithm was designed and successfully tested on biomedical images taken from the ADNI (Alzheimer's Disease Neuroimaging Initiative) dataset.
• The encryption algorithm was developed using a new chaotic system based on hyperbolic sine, and the key stream was created with this system.
• This structure ensured that the system was strong in terms of confidentiality, integrity, and legal compliance.
• The images were converted into one-dimensional vectors and passed through a two-stage diffusion process, thus creating a plaintext-dependent encryption key sequence.
• By choosing the Jerk (third-order derivative-based) chaotic circuit model, the circuit's realizability in an electronic environment was ensured.
• In this context, circuit simulations were performed with tools such as OrCAD-PSpice, and experimental verification was performed.
• The chaotic system used has a hyperchaotic nature with multiple positive Lyapunov exponents and exhibits complex dynamic behaviors such as multistability (multiple stable states), symmetry breaking, and bifurcation branching.
• The hyperbolic sinusoidal current-voltage characteristic obtained via antiparallel semiconductor diodes has increased the security and randomness of the encryption algorithm.

• I worked in the academic research part of the research project established by Dr. Lecturer Görkem SERBES for the following purposes:
• Participating in research on the classification of signals obtained with Doppler Sound.
• Using raw data converted into more significant on signal processing with Matlab (Hilbert Transform).
• Each processed signal was converted into a spectrogram image using the Short Time Fourier Transform.
• The architecture we developed combines ten different pre-trained Convolutional Neural Network models with a combination of Transfer Learning and Freezing Layer approaches.
• In the model combination process, Soft and Hard Voting methods were used as the Ensemble Learning approach.
• In this study, 400 Doppler ultrasound signals were analyzed to classify three categories: Speckle, Artifact, and Embolic.
• The signals were converted into spectrogram images using Short-Time Fourier Transform (STFT).
• A novel ensemble architecture was developed by fusing ten pre-trained Convolutional Neural Network (CNN) models through Transfer Learning and Freezing Layer strategies.
• Both Soft and Hard Voting methods were applied for ensemble learning.
• The model achieved a classification accuracy of 96.73% and an F1-score of 96.5%, demonstrating its high potential in the early and reliable detection of embolic particles-critical in stroke prevention.

• Aiming at solving health problems using engineering principles and methods.
• In this regard, the purpose of my internship is experience in maintenance and repair of eye hospital devices.
• Opportunity to contribute to the daily operation of eye hospital devices.
• Gaining knowledge and experience about eye health technologies.
• In this place, where I did my company internship, I gained knowledge and experience about the repair, calibration and working principle of the MS (mass spectrometer) device.
• I learned about the tuning and production of kits.
• And I also learned about the STM32 embedded system used in these devices, its basic operation and its effect on the devices.
• I gained 2 months of experience on the construction of three-phase electric motors and the design of DC three-phase motors.
• I also gained 1 month of experience on the design of the speed control card for the speed control of these devices and the use of the ESC module to control it using Arduino.

In this place, where I did my company internship, I gained knowledge and experience about the repair, calibration and

working principle of the MS (mass spectrometer) device. I learned about the tuning and production of kits. And I also

learned about the STM32 embedded system used in these devices, its basic operation and its effect on the devices.

I gained 2 months of experience on the construction of three-phase electric motors and the design of DC three-phase

motors. I also gained 1 month of experience on the design of the speed control card for the speed control of these

devices and the use of the ESC module to control it using Arduino.