Resumé

Resumé

You can find my Resumé here.

Summary

Highly accomplished and innovative Machine Learning Engineer with a strong background in Computer Science and Computer Engineering. Possessing a proven track record of developing cutting-edge machine learning solutions, optimizing deep learning models, and leading successful research and development projects. Excels in creating robust CI/CD pipelines, integrating advanced generative models into existing applications, and achieving exceptional performance in image-based generative services. A published researcher and award-winning participant in various prestigious competitions.

Education

2023 - Current Virginia Tech PhD in Computer Science

2018 - 2021 Boğaziçi University MSc in Computer Engineering

2012 - 2017 Boğaziçi University BSc in Computer Engineering

Publications

2024 CLoRA: A Contrastive Approach to Compose Multiple LoRA Models. arXiv preprint arXiv:2403.19776.

2024 Conditional Information Gain Trellis. arXiv preprint arXiv:2402.08345

2023 CONFORM: Contrast is All You Need For High-Fidelity Text-to-Image Diffusion Models. arXiv preprint arXiv:2312.06059. Accepted to CVPR 2024

2022 Unsupersived Routing Strategies for Conditional Deep Neural Networks. MSc Thesis. Boğaziçi University.

2020 BURST: Software and Simulation Solutions of an Autonomous Vehicle, 2020 28th Signal Processing and Communications Applications Conference

2018 Privacy-Preserving Intersection Management for Autonomous Vehicles, Proceedings of the Tenth International Workshop on Agents in Traffic and Transportation

Work Experience

Aug 2023 - Now Research Assistant, Virginia Tech

• Leading the research efforts on text-to-image generation.
• Contributed to the development and publication of methods enhancing the text-image fidelity of diffusion-based text-to-image models.
• Collaborated with Google to implement research findings in closed-source diffusion-based image generation models, resulting in a substantial improvement in image fidelity.

Nov 2022 - Aug 2023 Machine Learning Engineer, Lyrebird Studio

• Developed and maintained image-based generative machine learning services, processing and successfully handling an average of 5 million requests per day.
• Architected robust machine learning CI/CD pipelines using GitHub Actions and utilized AWS CDK for build architecture as code, enabling the seamless deployment of research team outputs as production-ready services.
• Implemented golden AMIs for GPU-accelerated instances using Packer, resulting in a remarkable reduction in boot-up time from tens of minutes to a few seconds, ensuring faster scale-up to meet high-volume demands.
• Led the design and deployment of stable diffusion-based model training and image generation services in the production environment, effectively handling thousands of requests per day on GPU-accelerated instances with high performance and stability.
• Integrated deep learning-based generative solutions into existing applications, significantly enhancing their capabilities and user experience.

Aug 2021 - Nov 2022 Machine Learning Engineer, Vispera

• Spearheaded the automation of deep learning model training using Python and TypeScript, resulting in a tenfold increase in daily model deployments, significantly reducing development time and costs.
• Led the launch of a user-friendly VueJS front-end service for the operations department, empowering researchers to efficiently train and deploy new models by providing real-time monitoring of online and offline metrics, enhancing model observability and researchers’ productivity.
• Successfully coordinated the transition of the deep learning stack to TensorFlow 2, streamlining the adoption of state-of-the-art deep learning models for production, leading to improved performance and maintainability.
• Worked as a full-stack machine learning engineer, using VueJS for frontend services; Python for machine learning services; TypeScript, NodeJS, Go, PostgreSQL, MongoDB, and GraphQL for backend services; Docker and Argo Workflows for containerization and orchestration.

Oct 2019 - Aug 2021 Computer Vision Research Engineer, Vispera

• Led research and development for deep learning image recognition models, utilizing Python, TensorFlow, and OpenCV, to solve challenging problems related to out-of-distribution recognition and hierarchical classification.
• Successfully implemented state-of-the-art deep learning image recognition models, achieving exceptional classification accuracy above 95% on online measurements, ensuring the delivery of high-performance solutions to meet business requirements.
• Pioneered the formulation and implementation of a novel zero-shot learning-based image recognition model using PyTorch, which significantly reduced image annotation time by four times. This innovative approach recommends best matches without annotated data, optimizing the model development process.

Aug 2018 - Oct 2019 Computer Vision Research Engineer, İdea Technology Solutions

• Introduced a novel tree-based deep learning architecture and method based on sparse execution of neural networks using Python, TensorFlow, and TensorFlow Lite for a project funded by The National Scientific and Technological Research Institution.
• Proposed a k-centroids-based clustering algorithm to determine better anchor boxes for object detection models, increasing the model’s object detection performance by approximately 15%.

Awards

2021 Winner, Teknofest RoboTaksi Autonomous Vehicle Competition - The Most Original Software Prize

2020 Runner-up, Anadolu Sigorta Datathon Challenge

2020 Finalist, Teknofest RoboTaksi Autonomous Vehicle Competition

2020 Winner, Teknofest RoboTaksi Autonomous Vehicle Competition Simulation Phase

2019 Finalist, Teknofest RoboTaksi Autonomous Vehicle Competition

2019 Winner, Teknofest RoboTaksi Autonomous Vehicle Competition Simulation Phase

2018 Winner, Mercedes-Benz Hack.Istanbul Hackathon

2017 Winner, BSH Analytics for Production Excellence Hackathon

2017 Runner-up, Boğaziçi University Computer Engineering Senior Projects Competition

2016 Finalist, TUBITAK Undergraduate Software Project Competition

Additional Experiences

2021 Guide, inzva - METU ImageLab AI Labs Joint Program

• Gave a lecture on probability, statistics, and graphical models for the Deep Generative Models. The course was organized in coordination with Gökberk Cinbiş from METU.

2018 - 2021 Head of Autonomous Vision Team, BURST (Boğaziçi University Autonomous Electrical Vehicle Team)

• Founded a team and laboratory for building an electric autonomous vehicle, creating autonomous driving research and development facilities at Boğaziçi University.
• Created a simulation environment using Gazebo, ROS, C++, and Python to simulate competition scenes, attaining the highest scores for two consecutive years in the National Autonomous Vehicle Competition.