Deniz Yener

Deniz Yener

Research & Development · AI for Life Sciences

Proactive professional focused on Research and Development for the Life Sciences. During my Erasmus Mundus master’s in AI for Sustainable Societies, I developed a deep learning–based, multi-resolution seizure detection model using multi-channel electroencephalography (EEG).

“Deniz worked on projects at the intersection of data science, artificial intelligence, and EEG for clinical applications in neurology. She is a serious, curious, and highly energetic professional who approaches research with both rigor and enthusiasm. Deniz quickly engages with complex topics and demonstrates strong motivation to learn and contribute. It has been a real pleasure working with her, both professionally and personally, and she brings a very positive and collaborative spirit to the team.”

Download CV

A Multi-Resolution Deep Learning Framework for Seizure Detection in Multi-Channel Electroencephalography

Deep Learning based multi-resolution seizure detection using multi-channel EEG

My Master of Science in Engineering thesis, completed within the Erasmus Mundus Joint Master's programme in AI for Sustainable Societies, develops a Deep Learning based multi-resolution seizure detection model using multi-channel Electroencephalography (EEG). The model learns temporal and spectral patterns across multiple resolutions to distinguish seizure from non-seizure activity in clinical EEG recordings.

Key achievements:

  • Designed a multi-resolution deep learning architecture for seizure detection from multi-channel EEG signals.
  • The introduced framework independently analyzes EEG segments of varying epoch lengths and integrates their predictive outputs to formulate a final diagnostic decision.
  • We preprocessed raw EEG signals and segmented them into varying non-overlapping temporal lengths to verify the impact of EEG data length on seizure detection performance.

Keywords: Epilepsy, Electroencephalography, Deep Learning, Seizure Event Detection, Cross-Subject Analysis, Decision Fusion

Nasa Earth Observation Hackathon

NASA Data Center

Kartta is a dashboard that helps urban planners, city governments, and policymakers understand the environmental and social risks associated with the construction and operation of data centers in cities.

Key achievements:

  • We are a data-driven platform that uses a wide range of environmental and urban datasets to generate a comprehensive overview of cities and human settlements.
  • By combining NASA satellite observations with local information (water resources, energy supply, population density, and household consumption patterns), Kartta creates an integrated picture of how urban areas function and evolve.
  • This data fusion allows planners and decision-makers to make more accurate, evidence-based decisions about where to locate new data centers and how to improve existing urban areas affected by current ones.
  • In short, we turn complex, multi-source data into clear insights for sustainable urban planning related to data center constructions and maintenance.

Technologies used: Python, Streamlit, QGIS, Figma, Scikit-learn, Pandas, NumPy

AI Adoption Readiness Assessment

AI Adoption Readiness Assessment

A comprehensive AI Adoption Readiness Assessment designed for Coresbond to evaluate the organization's preparedness for integrating AI technologies. This assessment covers key areas such as data infrastructure, talent capabilities, ethical considerations, and strategic alignment to evaluate the organization's readiness to ensure successful AI implementation.

Key achievements:

  • The 95% MIT report shows that the "vast majority" of organizations, including SMEs, "remain stuck with no measurable P&L impact" from their AI initiatives (Challapally et al., 2025).
  • A structured questionnaire was employed to elicit feedback from leading professionals regarding the tool's efficacy in assessing AI adoption and usage efficiency.
  • By providing a systematic assessment of strategic maturity and regulatory viability, the tool equips SMEs to transition from exploratory, peripheral AI use to targeted, core operational integration, thereby maximising the potential for measurable financial returns and ensuring legal operational resilience in a regulated environment.

Technologies used: TypeScript, CSS, HTML, Json

Voice Enabled Medical Chatbot

A Voice-Enabled Healthcare Chatbot for Neurodegenerative Disorder Identification

This project is a voice-enabled healthcare chatbot that utilizes deep learning to assist users in identifying potential neurodegenerative disorders based on their symptoms. The chatbot supports both female and male voice options and interacts in natural language conversations to provide insights.

Key achievements:

  • Natural language processing (NLP) for symptom analysis.
  • Voice interaction (male/female options).
  • Deep learning model trained to recognize patterns in neurodegenerative disorders.

Technologies used: Deep Learning, NLTK, WordNetLemmatizer, Keras, Dropout, Activation Function, SGD Optimizer

Heart Disease Prediction Model

Heart Disease Prediction

Developed a machine learning model to predict heart disease risk using patient data. This project focuses on creating a reliable and accurate prediction system that can assist healthcare professionals in early diagnosis.

Key achievements:

  • Achieved 92% accuracy in heart disease prediction
  • Implemented multiple ML algorithms for comparison
  • Created an intuitive user interface for predictions
  • Optimized model for high recall to minimize false negatives
  • Conducted thorough feature importance analysis

Technologies used: Python, Scikit-learn, Pandas, NumPy, Matplotlib

Beijing PM2.5 Air Quality Prediction

Beijing Air Quality

Developed a machine learning model to predict PM2.5 air quality levels in Beijing using historical data from 2010-2014. After forecasting PM2.5 air pollution levels in Beijing, evaluated whether meteorological variables alone could predict pollutant concentrations effectively without engineered lag and rolling features.

Key achievements:

  • Built predictive models achieving 95% accuracy using advanced ML algorithms
  • Implemented interactive data visualizations for trend analysis
  • Analyzed 5 years of environmental data with 438,000+ data points
  • Identified key correlations between weather conditions and PM2.5 levels
  • Developed a robust data preprocessing pipeline for handling missing values

Technologies used: Python, Pandas, Scikit-learn, Plotly, Winsorizer, Streamlit