✖
Ömer Tuğrul

1class ContactInformationCard:
2 def __init__(self):
3 self.dept = "EEE @ Bilkent University"
4 self.lab = ""
5 self.email = "o.tugrul@ug.bilkent.edu.tr"
6 self.phone = "+90 534 646 09 42"
7
8 def flipCard(self):
9 print("tap on the card to flip.")
10
11 def closeCard(self):
12 print("tap outside to close it.")

Ömer Tuğrul

Ömer Tuğrul

Senior Electrical and Electronics Engineering Student at Bilkent University.

I am currently focusing on building efficient AI agents with hands-on experience in deep neural networks. Also I am developing chatbots with sophisticated RAG pipelines.

LLMs: AI agents, RAG pipelines
GANs: New GAN architectures

📜 updates

I am happy to share that I started NLP Engineer Intern @ Databoss A.Ş.
July 2024

📚 publications

SWaTGAN Sliding Window Attention for Efficient Image Generation
CS485 Term Project Paper
Ömer Tuğrul, Mennan Gök
[PDF] | [CODE]
Abstract: IN this project, we propose the novel architecture Sliding Window Attention-GAN (SWaTGAN), which utilizes the sliding window attention (SWA), proposed in TransGAN. SWA limits the receptive field of each token to a pre-determined window size. In this way, in the early layers of transformer blocks, the similarity score is calculated between the near-by tokens. The receptive field of each token increases towards the latter transformer layers, enabling the model to capture long-range dependencies despite starting with a narrow receptive field. ... See More

🧩 outreach activities

Bird Species Classification

Ömer Tuğrul, Mennan Gök, Selin Ataş, Gökay Balcı
[POST] | [CODE]
Description: The Bird Species Classification Project, a term project for the Cs464- Introduction to Machine Learning course, is an academic initiative that explores the synergy between avian biodiversity and computational intelligence.
CycleGAN

Ömer Tuğrul
[POST] | [CODE]
Description: Unofficial Pytorch implementation of ”Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks” paper.

⛏️ resources

all software releases of the above projects can also be found here!

‎