Close

Cavit Cakir


 Resume  Github   Linkedin

About Me

As a recent Master’s graduate in Informatics, I bring a comprehensive background in Computer Science, with specialized expertise in Computer Vision, Natural Language Processing (NLP), Machine Learning, and Artificial Intelligence (AI). My academic pursuits have been complemented by hands‑on experience in programming languages such as Python and C++, and proficiency in utilizing modules like PyTorch, Numpy, Pandas, and TensorFlow for developing advanced AI solutions. My professional experience spans working in the domains of Computer Vision and NLP across two different companies, where I applied my skills to solve real‑world problems. My master’s thesis, which focused on autonomous driving through computer vision, further underscores my capability to tackle complex challenges in AI. As I seek my first full‑time opportunity, I am eager to bring my unique blend of academic knowledge, practical experience, and passion for technology to make a meaningful impact in the field.

Education

Technical University of Munich (TUM)

Oct 2021 ‑ Mar 2024 (Expected)

Master of Science in Informatics

Sabanci University

Sep 2016 - Jun 2021

Bachelor of Science in Computer Science and Engineering

Experience

Computer Vision Engineer Working Student

Quasara GmbH, Munich/Germany

  • Led the end-to-end development of a Damage Classification project, from initial data preprocessing to model training and implementation, significantly enhancing image analysis performance by utilizing state-of-the-art transformer models. This role showcased not only strong project ownership and technical expertise but also proficiency in applying cutting-edge AI technologies in practical applications.

  • Facilitated communication with the client company through regular updates, in-depth discussions on findings, and comprehensive presentations of model outcomes, demonstrating exceptional client engagement and presentation skills.

Natural Language Processing (NLP) Intern

FineSci Technology, Istanbul/Turkey

  • Contributed to a News Classification and Clustering project by focusing on the classification of news using NLP techniques. This involved detailed work on data preprocessing, model selection, and tuning to enhance the accuracy and efficiency of news categorization.

  • Utilized state-of-the-art transformer language models to classify news articles, showcasing expertise in applying cutting-edge NLP technologies. This role demonstrated the ability to leverage advanced algorithms for practical applications, significantly improving project outcomes through innovative solutions.

Undergraduate Teaching Assistant

Sabanci University, Istanbul/Turkey

  • Independently managed lab sessions for 20 - 30 students, providing hands-on support and guidance, alongside offering personalized mentoring during office hours to enhance students’ understanding and engagement with course materials.

  • Utilized C++ as the primary programming language to facilitate practical learning experiences in the Introduction to Computing course, demonstrating deep technical knowledge and application skills.

Selected Projects in Computer Vision

Self‑Supervised Feature Learning for 3D LiDAR Semantic Segmentation with Neural Radiance Fields (NeRFs) at CARIAD

Master’s Thesis     |     Python, Pytorch, Transformers, NeRF     |     May 2023 - Dec 2023

  • Conducted under CARIAD, the automotive software company of Volkswagen, focusing on the autonomous driving domain.
  • Managed the entire project lifecycle, from data preprocessing to the novel application of Neural Radiance Fields (NeRFs) and 3D LiDAR models, demonstrating a comprehensive approach to complex challenges in the field.
  • Utilized state‑of‑the‑art techniques to significantly reduce the reliance on labeled datasets in autonomous driving, demonstrating the potential for more efficient and scalable model training methodologies.
View Thesis

Implementation of Panoptic Neural Field at TUM

Advanced Practical Course     |     Python, Pytorch, NeRF     |     Oct 2022 - Mar 2023

  • Implemented the Panoptic Neural Field (PNF) paper from scratch using Kaolin Wisp and the KITTI 360 dataset, demonstrating initiative and expertise in translating theoretical concepts into practical, executable code in the absence of pre-existing implementations.
  • Optimized the PNF architecture for enhanced performance, employing advanced techniques to achieve significant improvements in model efficiency and accuracy, showcasing my ability to innovate and improve upon existing methodologies.
View Project

3D Perception for Autonomous Driving Survey at TUM

Advanced Seminar Course     |     Research     |     Apr 2022 - Sep 2022

  • Conducted in-depth research on 3D Object Tracking Methods, with a particular focus on utilizing infrastructure sensors. This involved a comprehensive comparison of cutting-edge methodologies presented in recently published papers from top-tier conferences, demonstrating my analytical skills and ability to synthesize complex information.
  • Written a survey paper that encapsulates the findings and insights derived from comparing recent advancements in 3D object tracking.
View Project

Improving Point Cloud Transformer using Curve Aggregation at TUM

Machine Learning for 3D Geometry Course Project     |     Python, Pytorch, Transformers     |     Apr 2022 - Sep 2022

  • Improved the shape analysis capabilities of a Point Cloud Transformer by implementing a Curve Aggregation method, showcasing my ability to enhance model performance for detailed 3D point cloud processing. This novelty led to more accurate shape interpretations on the ShapeNet Parts dataset.
  • Successfully converted the Point Cloud Transformer’s implementation from Jittor to PyTorch, demonstrating technical proficiency and versatility in adapting complex models to widely used frameworks.
View Project

Skin Cancer Classification at Sabanci University

Machine Learning Course     |     Python, Tensorflow     |     Feb 2020 - Jun 2020

  • Used transfer learning techniques to facilitate early diagnoses of skin cancer through image analysis of skin segments. This approach enhanced the accuracy and reliability of diagnostic processes.
View Project

Selected Projects in NLP

NLP and Knowledge Graphs for Research Cluster Prediction and Analysis at TUM

TUM-DI-LAB Interdisciplinary Project     |     Python, Pytorch, Transformers     |     Oct 2022 - Mar 2023

  • Participated in Unsupervised Classification of Research Papers Project and proposed a novel Hierarchical Classification Method while successfully applying existing methods.
  • Used the cutting-edge SPECTER embedding model to enhance the quality of embeddings substantially. This choice enabled a more precise and nuanced understanding of research paper content, contributing to the project’s success.
View Project

Emotional Clustering of Social Media Users at TUM

Advanced Practical Course     |     Python, Pytorch, Transformers, BERT     |     Apr 2022 - Sep 2022

  • Used the pre-trained BERT model, extracted embeddings to cluster users based on their textual data. This approach demonstrated my proficiency in utilizing advanced NLP techniques for user segmentation.
  • Preprocessed Reddit users’ posts to align with the pre-trained BERT model’s requirements, showcasing my ability to prepare and adapt large-scale datasets for complex NLP tasks.
  • Applied a variety of Dimensionality Reduction Methods, including HDBSCAN and KMeans, to analyze and cluster high-dimensional data efficiently.
View Project

Meeting Scheduler Chatbot at Sabanci University

Bachelor’s Graduation Project     |     Python, JavaScript, React, Node.js, Docker     |     Sep 2020 - Jun 2021

  • Developed an advanced chatbot utilizing the RASA Bot Framework, incorporating pre-trained natural language understanding methods to accurately process and interpret user inputs. This approach showcased my ability to use state-of-the-art AI technologies for creating interactive and intelligent conversational agents.
  • Implemented a graphical user interface for the chatbot using React and Node.js, encapsulated within Docker containers for ease of deployment and scalability. This end-to-end development highlights my proficiency in full-stack development and my commitment to delivering robust, scalable applications.
View Project

Skills

Get in Touch