E

Eray Aydemir

AI Developer | Full Stack Developer

Full Stack Developer and Machine Learning Engineer specializing in deep learning, NLP, and cybersecurity. Conducting cutting-edge research at WiseLab@Hacettepe on detecting malicious functions from reverse-engineered malware samples using LLM models.

About Me

Background

Currently pursuing MSc in Information and Data Engineering at Hacettepe University (GPA: 3.68). Working as an AI Developer at Kloudser, developing full-stack applications and ML solutions. Conducting graduate research at WiseLab@Hacettepe on malware detection using LLM models to identify malicious functions from reverse-engineered malware samples.

Research Focus: Leveraging Large Language Models for cybersecurity applications, specifically detecting malicious code patterns in decompiled malware binaries.

Location: Ankara, Turkey

Technical Skills

Programming Languages

Python JavaScript TypeScript Java Rust C

ML/AI Frameworks

PyTorch Scikit-learn XGBoost LLM Hugging Face

Web Development

FastAPI SvelteKit Spring Boot Node.js

Tools & Cloud

Docker AWS Git PostgreSQL Linux

My Projects

A collection of my work in web development and machine learning

LLM-Based Malicious Function Detection

ML/DL

Graduate research project using Large Language Models to detect malicious functions in reverse-engineered malware samples. Developing novel approaches for analyzing decompiled code patterns.

LLMReverse EngineeringPyTorchHugging FaceCybersecurity

Personal Portfolio Website

Web Dev

Modern, responsive portfolio website built with SvelteKit and Tailwind CSS. Features project showcase and contact form with FastAPI backend for form handling.

SvelteKitTypeScriptTailwind CSSPythonFastAPI

Tuvens - Event Curation Platform

Web Dev

Full-stack event curation platform enabling users to discover, create, and manage events. Built with SvelteKit frontend, NestJS backend, AWS RDS PostgreSQL database, Cognito authentication, and S3 for storage.

SvelteKitNestJSNode.jsPostgreSQLAWS CognitoS3Tailwind CSS

Strand345 - Dynamic Website

Web Dev

Dynamic website with interactive pages and integrated contact form. Built with SvelteKit, Tailwind CSS, Python backend, and Google Cloud API integration.

SvelteKitTypeScriptTailwind CSSPythonGoogle Cloud API

NGO Hukuk - Legal Blog

Web Dev

Professional legal blog platform with content management system, article publishing, and SEO optimization. Built with Next.js and Tailwind CSS frontend, FastAPI backend, PostgreSQL database, deployed on VPS with Docker and Nginx.

Next.jsTailwind CSSPythonFastAPIPostgreSQLDockerNginx

GeoID - Business Website

Web Dev

Professional business website with integrated contact form. Built with SvelteKit and Tailwind CSS featuring dark/light theme toggle for enhanced user experience.

SvelteKitTailwind CSSDark/Light ThemesForms

LLM-Powered Android UI Agent

ML/DL

Designed Python agent integrating fine-tuned LLM with adb and emulator for autonomous UI exploration and interaction.

PythonLLMAndroidADB

Android Malware Detection

ML/DL

Built and deployed calibrated XGBoost model on 200K×44K dataset achieving 98% validation accuracy. Calibration ensures reliable probability outputs for better decision-making. Used reverse engineering and static analysis for feature extraction.

XGBoostCalibrationScikit-learnFastAPIDocker

Android Malware Type Classification

ML/DL

Multi-class classification system using calibrated XGBoost to identify malware types including Trojan, Downloader, Dropper, and other variants. Achieved 94% accuracy with calibrated probability outputs for reliable predictions.

XGBoostCalibrationClassificationPythonMalware Analysis

Binary Text Classification for Suicide Detection

ML/DL

Compared LSTM with BERT embeddings and TF-IDF-based neural networks achieving ~93% accuracy for suicide risk detection.

LSTMBERTNLPPyTorch

Android Permissions Analyzer

ML/DL

Extracts Android apps' permissions and explains to users what the app is capable of using fine-tuned Mistral 8B model.

Mistral 8BFine-tuningAndroidNLP

Windows Malware Classification

ML/DL

Implemented CNN model on executable-to-image transformations for malware research and classification.

CNNDeep LearningPythonImage Processing

© 2025 Eray Aydemir. All rights reserved.

Built with SvelteKit & Tailwind CSS