Deep learning model for predicting brain age from EEG data. Features include preprocessing pipeline, connectivity analysis, and age prediction using neural networks.
Alp Akova
PhD Candidate in Cognitive Brain Sciences
Researcher in Cognitive Aging, EEG, and Machine Learning
About Me
I am a PhD candidate at the University of Trento, specializing in cognitive aging, EEG-based biomarkers, and machine learning applications in neuroscience. My research focuses on understanding how physical activity influences brain connectivity and cognitive health in aging populations, leveraging EEG-based age prediction (brain-PAD) and advanced machine learning models to explore neural aging trajectories.
My work integrates EEG connectivity analysis, neural synchronization, and frequency-specific oscillations to identify markers of cognitive resilience. I employ deep learning, ensemble methods, and multimodal data integration to enhance the accuracy of brain age prediction models.
Beyond my doctoral research, I have experience in large language models (LLMs), NLP, and decision-making experiments, having worked at Subconscious AI and the University of Padova's Baby Lab. My expertise spans EEG analysis, signal processing, graph theory metrics, and data-driven cognitive modeling.
Research
Current Research
My primary research investigates the relationship between physical activity and cognitive aging, using EEG-based brain age prediction (brain-PAD) and advanced machine learning techniques. This work aims to develop personalized interventions for promoting brain health in aging populations.
Research Interests
- Cognitive Aging and Brain Health
- EEG Analysis and Brain Age Prediction
- Machine Learning in Neuroscience
- Physical Activity and Neural Plasticity
- Neural Synchronization and Brain Networks
- Large Language Models and NLP
Keywords
Completed Projects
Code repository replicating research on using LLMs to simulate human behavior in experimental settings. Features implementations of key papers exploring AI as participant substitutes in psychological research, with prompts and processing code for reproducing results at small scales.
Full-stack application for speech-to-text transcription using Apple's MLX framework with the Whisper model, optimized for Apple Silicon. Features a FastAPI backend for handling transcription requests and a Streamlit frontend that allows users to upload audio files, configure settings, and download transcripts in various formats.
Developed a tool that uses local LLMs via Ollama to automatically generate language learning materials. Creates reading passages tailored to specific topics, languages, proficiency levels, and styles, then generates relevant multiple-choice questions. Features an agentic workflow with quality control, text analysis, and interactive quiz functionality.