Merve Karalı
AI Researcher & Data Engineer
About me
I am an AI Researcher & Data Engineer, holding an M.Sc. degree in Computer Science from the Technical University of Munich. My research delves into the convergence of Deep Learning and Computer Vision, with a specialization in unsupervised and self-supervised learning, multi-modal data learning, and semantic segmentation.
In my professional journey, I’ve seamlessly navigated the realms of enterprise ML/AI and data engineering. As an AI Researcher, my contributions to projects like FunKI and 6G-SKY include exploring advanced image processing techniques, such as image super-resolution, restoration, and Trustworthy AI, with a particular emphasis on Explainable AI in computer vision. Bridging the research-production gap, I’ve actively engaged in projects automating ultrasonic testing analysis using CNNs and Transformers.
With eight years of experience as a Data Engineer, I’ve spearheaded initiatives in Data Warehouse, ETL, and Big Data across diverse sectors, including banking, telecommunication, e-commerce, and healthcare. My skill set encompasses proficiency in Python, SQL, Google Cloud Platform, and AWS.
Beyond the tech realm, I find joy in math, computing, engaging conversations, reading, cooking, hitting the gym, and more.
- Deep Learning
- Computer Vision
- ML Research
- Data Science
- Data Engineering
Msc in Computer Science, 2023
Technical University of Munich
Bsc in Computer Science & Engineering, 2016
Koc University
Experience
- Collaborated in a dynamic research environment, contributing hands-on expertise in Python.
- Engineered advanced deep learning architectures using PyTorch Lightning and TensorFlow.
- Worked on the FunKI research project, leveraging deep learning frameworks to explore and optimize 5G network performance.
- Implemented anomaly detection techniques for enhanced diagnostics in ultrasound images.
- Contributed significantly to the 6G SKY research project, focusing on the explainability, interpretability, and robustness of deep learning networks.
- Spearheaded the replacement/migration of relational databases to cloud infrastructure (AWS, GCP).
- Executed tasks in DWH, ETL, and Big Data domains.
- Implemented and optimized end-to-end ETL processes, emphasizing extraction, CDC, load, recycle management, dimension management, and key management.
- Enhanced frameworks for classical DWH architecture, improving efficiency with large datasets.
- Assisted in documentation, analysis, quality assurance, testing, and operational aspects of ETL processes.
- Led the ETL workflow and processes as part of the DWH Reengineering project, designing and developing DevOps CI/CD pipeline and migration flows.
- Developed productivity tools and automation, reviewed designs and code, and contributed to data preparation for datamining.
- Leveraged cloud services on AWS and GCP to gather data from OLTP systems, ensuring scalability and efficiency.
- Employed cloud-based optimization techniques on AWS and GCP to improve the efficiency and performance of database queries and processes.
- Managed ETL architecture, optimizing data processing for operational data.
- Improved database performance through property optimization and mapping enhancements on ODI.
- Developed finance applications using RDBMS systems (Oracle Database, IBM DB2) and Java for efficient data manipulation in multichannel communication.
Projects
This project focuses on detecting defects in jet engines using ultrasound images, where the involved tasks included analyzing varied input data, preprocessing different data modalities, and implementing advanced techniques like anomaly detection, self-supervised learning, and transfer learning for machine vision applications in a military context.
The project aimed to investigate and implement state-of-the-art methods to enhance the explainability, interpretability, and robustness of deep learning networks in the context of 6G Sky applications.
The project aimed to assess the efficacy of data-driven machine learning methods, particularly deep learning models, for enhancing channel estimation in 5G communication systems using FunKI.
The project involves actively participating in a laparoscopic cholecystectomy operation to address and resolve the depth perception challenge associated with this surgical procedure.
The project aims to predict gene expression by analyzing millions of random promoter sequences, addressing challenges such as class imbalance and comparing the efficacy of convolution-based and attention-based networks in capturing local patterns.
The goal of this project is to predict prostate cancer by implementing the learning pipeline with self-supervised learning and transfer learning.
