Where Do LLMs Still Struggle? An In-Depth Analysis of Code Generation Benchmarks
IEEE/ACM International Conference on AI-powered Software (AIWare 2025)
Hello, I'm
Artificial Intelligence · Computer Science · Music
I'm an MSc. student in Artificial Intelligence and Machine Learning at TU Darmstadt, focusing on NLP & Software Engineering, Multimodal AI, and Deep Generative Models. Currently a HiWi at the Software Technology Group, working on LLM code generation benchmarks. Previously a research assistant at the Cosmos AI Lab at Yıldız Technical University.
Technical University of Darmstadt · Darmstadt, Germany
Notable courses: Multimodal AI, NLP for Web, Deep Generative Models, Computer Vision, Scalable Data Management Systems, AI for Coding Assistance.
Fontys University of Applied Sciences · Eindhoven, Netherlands
Exchange semester in international software engineering environment focusing on Agile practices, DevOps, and full-stack development.
Yıldız Technical University · Istanbul, Turkey
Honored Student. Bachelor's thesis: Code Generation with LLMs.
Technical University of Darmstadt · Darmstadt, Germany
Security-aware Code Generation and evaluation, and Benchmarking code generation with LLMs; data analysis, dataset preparation and experiment automation.
Cosmos AI Research Group, YTU · Istanbul, Turkey
Assisted model training, data pipeline development, and RAG system implementation.
IEEE/ACM International Conference on AI-powered Software (AIWare 2025)
Springer / EAI Conference Proceedings
Things I do in my spare time
Prizmat · Ableton · DJ · TouchDesigner
Under the alias PRIZMΔT, I produce and mix electronic music, hypnotic groovy techno and funky house. On the side, I build audio-reactive visuals for DJ/VJ sets and gesture-controlled instruments using TouchDesigner.
Bouldering · Running · Weightlifting · Cycling
I try to move in as many ways as possible, bouldering for problem-solving with my four limbs, running and cycling to be cardially healthy, weightlifting to stay grounded. Ideally all in the same week every week.