Mustafa Mert Tunalı

Software engineer focused on computer vision, large language models and ml infrastructure, strong interest in bio-inspired AI and vision transformers.

Proficient in Python, Go and TypeScript with experience using PyTorch, TensorFlow, HuggingFace, LangChain, LMQL, Docker, gRPC, Kubernetes and AWS. Experienced in optimizing deep learning models for efficient inferencing. Strong interest in machine learning infrastructure and bio-inspired AI models.

My work primarily focuses on the development and deployment of advanced models for tasks such as image analysis, classification, segmentation, and object detection. In addition, I also specialize in natural language processing (NLP) and the application and optimization of large language models. Specifically, I work on adapting these language models for various use cases, enhancing their performance, and implementing these models in practical applications.

Email  /  Google Scholar  /  Twitter  /  Github  /  LinkedIn

Analysis of Algorithms, Data Structures, Object Oriented Programming, Database Managment Systems, Linear Algebra, Calculus 1-2, Operating Systems, Computer Architecture, Computer Networks, Software Engineering, Discrete Mathematics, Probability and Statistics, Digital System Design.

This project originated from the need to seamlessly use OpenAI's API with custom features on the server-side and effortlessly stream this data to the frontend. It was designed to incorporate aspects like prompt engineering and rule-based systems. I believe this tool, born out of practical necessity, can assist others in similar projects and scenarios.

I aimed to make using pre-trained deep learning models simple and accessible through a user-friendly graphical interface, eliminating the need for additional coding and simplifying the setup process. Progress can be monitored in real-time using tools such as TensorBoard.

I'm interested in machine learning infrastructure and bio-inspired AI models. Much of my research is about transforming deep learning models into efficient and scalable systems.

This study aims to use deep learning to improve quality control in production lines by classifying steel surface defects using a limited data and computing power.