Experience

• Design and maintain production data pipelines over large-scale Turkish Airlines event data, transforming raw, high-volume data into reliable analytical and ML-ready datasets through automated backfills, idempotent processing, schema evolution, and performance optimization.
• Develop reusable Machine Learning and Explainable AI (XAI) workflows, combining leakage-safe feature engineering, model development and evaluation, automated feature auditing, and SHAP-based interpretation to understand the drivers behind complex customer and product behavior.
• Conduct statistical and data-science investigations that directly inform product and engineering decisions, using hypothesis testing, stratification, sensitivity analysis, and model-based diagnostics to distinguish genuine behavioral patterns from technical, measurement, and data effects.
• Build systematic data-quality and analytical frameworks for production environments, turning recurring validation, instrumentation, schema, and methodological checks into reproducible processes and documenting approaches so they can be consistently applied across projects and teams.

• Built and shipped a production-grade AI video analytics platform end-to-end, owning the full ML lifecycle from data collection and self-annotation to model training, evaluation, inference, GPU orchestration, deployment, and continuous optimization.
• Developed and fine-tuned the core computer-vision stack, training five task-specific models for ball/rim detection, court keypoints, action recognition, and jersey-number OCR using self-labelled data; designed custom tracking and player-identification logic to achieve robust results.
• Architected the complete MLOps and distributed inference pipeline, parallelizing full-match processing across GPU containers and integrating the ML backend with a production web application, database, storage, authentication, and deployment infrastructure.
• Optimized the platform for production-scale reliability and efficiency, redesigning the inference pipeline, GPU allocation, and candidate filtering to significantly reduce compute requirements while improving the accuracy and consistency of automated scoring analysis; validated the system on real-world matches across multiple countries and shipped it as a live product.

• Developed a time-series and econometric analysis pipeline for TÜBİTAK BİLGEM’s Mali Zeka fiscal-intelligence program, working with large-scale U.S. federal expenditure data to study dynamic and directional relationships between government departments.
• Applied a rigorous statistical methodology spanning stationarity testing, lag selection, Granger causality, and Vector Autoregression (VAR), including impulse-response analysis to characterize how expenditure dynamics propagate across departments and produce a framework transferable to public-sector fiscal data.

• Developed an end-to-end computer vision pipeline for an embedded system on NVIDIA Jetson Orin, combining real-time segmentation, tracking, and vision-based geometric estimation into a functional prototype.
• Benchmarked and optimized segmentation models for edge deployment, evaluating lightweight vision architectures with ONNX and TensorRT under strict latency and compute constraints, and developing geometry-based methods to improve measurement robustness.
• Contributed to applied research in autonomous perception and visual localization, evaluating modern localization approaches and emerging Vision-Language and World Models to inform architecture and technology decisions for future autonomous systems.

• Worked across data collection, processing, analysis, and integration, building Python-based workflows with Pandas and NumPy and connecting data sources through web scraping, APIs, and databases.
• Contributed to applied ML initiatives alongside data engineering work, including image classification and automated data-processing workflows for e-commerce applications.

• Used Python to analyze customer data and create visualizations, turning raw information into reporting that supported data-driven business decisions.
• Built reporting and process-automation workflows with Power BI and Power Automate, connecting data preparation, analysis, and recurring reporting tasks.
• Applied Robotic Process Automation (RPA) to streamline repetitive data-processing activities and improve the efficiency of operational workflows.
• Performed Recency, Frequency, and Monetary (RFM) analysis to understand customer behavior, segment customers, and support strategic business decisions.

• Developed data collection and organization tools for Amazon sellers, simplifying the process of gathering information and preparing it for day-to-day e-commerce operations.
• Designed user-friendly applications that made data workflows accessible to non-technical users, helping sellers work with information through straightforward interfaces.
• Built automation tools to streamline recurring e-commerce tasks, reduce manual work, and make operational processes more efficient and easier to manage.

Computer Vision Engineer · Oct 2022 – May 2023· 8 mos

• Worked on a TÜBİTAK-supported image-recognition and deep-learning project using a custom 100-class dataset, contributing across dataset preparation, model development, and evaluation.
• Prepared and manually labeled image datasets, applying preprocessing techniques such as histogram equalization to improve data quality and consistency before training.
• Experimented with CNN architectures including VGG16, VGG19, InceptionV3, and ResNet50, comparing their behavior and adapting model configurations to the project domain.
• Contributed to model-performance optimization and accuracy improvements through iterative experimentation and domain-specific customization.
• Worked on YOLO-based object detection and classification, contributing to dataset preparation, image preprocessing, and experiments that extended the project toward real-time recognition.

Junior Data Scientist · May 2022 – Oct 2022· 6 mos

• Built Python-based web-scraping workflows with BeautifulSoup and Selenium to collect and organize data from online sources.
• Developed SQL databases to structure the collected information and support efficient storage, retrieval, and further analysis.
• Analyzed scraped data to identify useful insights and used Google Firebase as part of the data-storage workflow.
• Created a lightweight Python graphical interface that made it easier to access, explore, and interact with the data-collection system.

Data Science Intern · Feb 2022 – May 2022· 4 mos

• Supported data-analysis and data-mining tasks, working with collected information to develop practical experience in data-science workflows.
• Assisted with web scraping and data collection, preparing information for subsequent processing and analysis.
• Contributed to SQL database development, helping organize and store data as part of the team's production-oriented workflows.