ÖZLEM TUNA

Designed and delivered specialized workshops on Python for data analysis tailored to life science professionals, covering NumPy, Pandas, data visualization, and reproducible workflows.

Prepared comprehensive hands-on exercises and course materials, emphasizing reproducible research practices, scripting best practices, version control, and advanced data-cleaning techniques.

Mentored numerous participants in applying advanced bioinformatics techniques to their unique datasets, providing individualized feedback to significantly strengthen their reproducible research workflows.

Developed and deployed 20+ real-world ML, deep learning, and generative AI projects (NLP, computer vision), delivering production-aligned models for classification, forecasting, and summarization.

Engineered and implemented LangChain-based AI agents and Retrieval-Augmented Generation (RAG) pipelines, significantly automating document summarization and customer-support functions.

Fine-tuned transformer models using LoRA and Hugging Face Transformers, establishing robust training and evaluation pipelines with PyTorch and MLflow for enhanced model performance.

Constructed comprehensive end-to-end data workflows for ingestion, cleaning, feature engineering, and validation using Python, Pandas, NumPy, and SQL, automating model evaluation and monitoring.

Analyzed extensive transcriptomic datasets, including 600+ bulk samples and over 150k single-cell profiles, successfully identifying candidate cancer biomarkers and molecular signatures.

Developed and deployed reproducible bioinformatics pipelines utilizing Nextflow and Docker for High-Performance Computing (HPC) environments, enabling scalable and portable analyses across diverse computational clusters.

Implemented advanced Cox proportional-hazards survival models, deriving critical risk scores to accurately stratify patient cohorts and support robust prognostic assessments for improved clinical insights.