Aybars Manav

• Developed a market-competitive 3D desktop scanner from scratch and currently developing an Intraoral Scanner (IOS) as the lead computer vision engineer. • Implemented and modified various structured light methods to solve the dense correspondence problem, implementations were done from scratch and on both CPU and GPU (using CUDA). • Implemented my own structured light method to be used in our new IOS, this method has proven itself to be superior to other well-known and real-time methods: It uses fewer patterns and is robust against subsurface scattering effects which are quite common for translucent objects (teeth, prosthesis, etc.). • Implemented several 2D and 3D filters that I designed by utilizing both my theoretical knowledge in signal processing and domain knowledge acquired by analyzing our specific problems. • Implemented many efficient mesh and point-cloud processing algorithms, some are implementations of papers and some are my own design. • Working on implementing a real-time image-based dense tracking and reconstruction algorithm using CUDA.

Worked on point cloud generation from images: • Implemented and modified structured light methods to solve the dense correspondence problem. • Implemented my custom control point detection algorithm on images. This algorithm was the product of fusing several papers in a smart manner. This method single-handedly improved OpenCV's calibration pipeline by more than 3 times. • Implemented a custom calibration pipeline. • Improved mentioned algorithms using domain-specific knowledge acquired by analyzing problems carefully and coming up with creative solutions.

• Modified and tuned the main surface reconstruction algorithm. • Implemented an efficient surface simplification algorithm.

• Worked in the Information Processing and Communications (IPC) lab led by Professor Deniz Gunduz. • Adapted SoTa ViTs for joint source and channel coding of images.(SWIN, Twins, Focal etc.) • Worked on fusing CNNs with ViTs for a hybrid architecture leveraging the inductive bias of CNNs. • Shared my architectures with the IPC lab as a library for the lab’s broader research. • Designed and implemented a hybrid code structure combining hand-crafted codes like QC LDPC and 5G Polar codes with learnable attention-based codes.

• Worked with Prof. Ali Emre Pusane on the project “Design and Implementation of Molecular Communication Systems Using Index Modulation” under the TUBITAK 2247-C Intern Researcher Scholarship Program (STAR). • Worked on a research project on mitigating Intersymbol Interference (ISI) in a SISO MCvD system and numerically proved an optimal reception interval exists.

•Worked on building a custom Turkish Open-domain Question Answering system by fine-tuning a BERT base transformer. • Building and evaluating different QA systems, different Information Retrieval and Reading Comprehension models. • Learned and implemented various NLP tasks for transformers such as Named Entity Recognition (NER), Part of Speech (POS) tagging, Sentiment Analysis, Text Classification, Long-form/Generative Question Answering, and Extractive Question Answering.

• Supervised, organized 60+ members of the student branch throughout various activities such as Summits, Workshops, Tea-Talks, Technical Courses, etc. • Organized Bogazici University Leading Women in Science and Technology (BTLK in Turkish) with a team of volunteers and 500+ participants, 15+ professional speakers. • Kept in touch with IEEE global, gave thorough monthly reports, participated in global, local events • Got in touch with professionals from various areas in order to get sponsorships, build partnerships, etc.

Social Responsibility Project for Education of Children. Purpose was to establish connections and bond between international volunteers and children of Romania to eradicate racism and widen people's ideas towards each other, learning different cultures and teaching English to Romanian children/students.

Received the prestigious Justus van Effen Excellence Scholarship (full tuition + monthly pay ~ 40K€/year) which is awarded to only two people in each faculty. GPA:9/10