Developed novel differentiable frameworks integrating combinatorial optimization algorithms into deep neural networks. Research focus: optimal transport theory, ranking-based losses, and advanced object detection architectures. Funded by TUBITAK 2247-C STAR National Scholarship Program.

Capstone Project: Designed and implemented a Transformer-based question-answering system utilizing T5 pre-trained architecture for extractive and abstractive text comprehension. Comprehensive curriculum spanning algorithms, systems programming, machine learning, and software engineering fundamentals.

Advanced research on continuous cybersickness detection in virtual reality environments using EEG-based multitaper spectrum estimation. Pioneered novel deep learning architectures achieving 30% performance improvement over baseline methods in real-time cybersickness prediction from brain signals, with potential applications in adaptive VR systems and user experience optimization.

• Developed learned frequency-weighted Kalman filters that denoise the innovation spectrum under spectral supervision
• Built EEG multitaper-spectrum deep models (ConvLSTM) for real-time VR cybersickness detection, +30% over baselines

Pioneering research on differentiable optimization frameworks for deep learning in computer vision. Developed novel methods for incorporating combinatorial optimization algorithms into neural network training pipelines, achieving state-of-the-art performance in object detection.

• Contributed to Bucketed Ranking Loss, speeding up ranking-based training of DETR/transformer detectors by up to 6x
• Developed optimal-transport (Sinkhorn-Knopp) and ranking-based loss formulations for end-to-end object detection

• Built a self-driving stack for a physical RC car, and trained RL agents for autonomous navigation task in ROS/Gym/PyTorch
• Simulated a Hector-quadrotor swarm in ROS/Gazebo, approximating numerical solvers for multi-agent localization

Developed sophisticated multi-modal deep learning systems for cardiac and sleep abnormality detection from physiological signals. Designed novel architectures integrating time-series analysis with spectral feature extraction for beat-by-beat cardiac arrhythmia classification, achieving competitive performance in international challenges. Published research findings at Computing in Cardiology (CinC) 2021.

• Built multi-modal deep models for beat-by-beat cardiac and sleep abnormality detection from physiological signals
• Fused time-series and spectral features for arrhythmia classification, competitive in international challenges

CNNFOIL: Approximating HLLC Riemann solver for flow prediction around airfoils via encoder-decoder neural networks. Engineered state-of-the-art neural network models achieving 4.4% performance improvement in predicting critical flow field parameters around airfoil geometries.

• Approximated HLLC Riemann solver for airfoil flow prediction via encoder-decoder networks
• Achieved 4.4% improvement in predicting flow field parameters around airfoil geometries

TUBITAK 3501 funded research project. Designed and implemented high-performance IDS pipeline optimized for containerized cloud infrastructures. Collaborated on Kubernetes misuse detection research, contributing to novel datasets and detection methodologies.

• Published at AINA 2024 on microservices security architectures
• Received Best Paper Award at IEEE/SRMC 2022 for empirical analysis of IDS approaches
• Co-authored ITU FET journal paper introducing a Kubernetes misuse detection dataset

Textual analysis for irony detection and research impact prediction. Implemented end-to-end neural pipelines integrating GCN and LSTM networks for analyzing citation patterns and textual content on Twitter datasets.

• Published at DaWaK 2021 on explainability in irony detection
• Developed graph-based models for information retrieval

Healthcare policy simulation for evidence-based decision making. Utilized simulation models focusing on mechatronics applications in healthcare. Developed GCN-based citation graph embeddings for healthcare policy papers.

• Built simulation models for healthcare policy decision making
• Developed GCN-based citation graph embeddings for healthcare policy papers

Conducted recitations, graded assignments, proctored exams, and provided office hours.

• CENG 280 - Automata Theory (Spring 2024, Spring 2025, Spring 2026)
• CENG 242 - Programming Language Concepts (Spring 2022, Spring 2023, Spring 2025, Spring 2026)
• CENG 223 - Discrete Computational Structures (Fall 2023, Fall 2024, Fall 2025)
• CENG 424 - Logic for Computer Science (Fall 2023, Fall 2025)
• CENG 460 - Introduction to Robotics for Computer Science (Fall 2024)
• CENG 213 - Data Structures (Spring 2024)
• CENG 382 - Analysis of Dynamical Systems (Spring 2023)
• CENG 334 - Introduction to Operating Systems (Spring 2022)

Helped TAs in recitations and lab sessions, also organized social sessions during covid'19.

• CENG 240 - Programming with Python for Engineers (Fall 2021)
• CENG 111 - Introduction to Computer Engineering Concepts (Fall 2020)
• CENG 230 - Introduction to C Programming (Spring 2019)

Merit-based national scholarship awarded to students contributing to TUBITAK-funded research projects. Supported graduate thesis on differentiable combinatorial optimization at ImageLab, METU.

For "Bachelor of Science in Computer Engineering".

Merit-based national scholarship awarded to students contributing to TUBITAK-funded research projects. Supported undergraduate research on container security and intrusion detection (CONTSEC) at METU.

Merit-based national scholarship awarded to students contributing to TUBITAK-funded research projects. Supported undergraduate research on healthcare policy simulation at Koç University School of Medicine.