Abdulkadir Nar
AI Research Engineer at Vyvo
Sivas, Türkiye Turkey
Summary
🤩
Rockstar🎓
Top SchoolAbdulkadir Nar is an AI Research Engineer with eight years of hands-on experience building and optimizing generative models, from Stable Diffusion and ControlNet to custom Text-to-Speech systems. He has shipped production-level diffusion features and performance improvements—integrating mechanisms like FreeU and optimizing inference with Flash Attention, Triton and compilation toolchains—to accelerate real-world applications. A core contributor at AdCreative.ai, he helped deliver acquisition-grade features including automated background replacement and ADLLM fine-tuning that reached GPT-4 level performance. His open-source work includes enhancing Hugging Face’s diffusers (tiling/canvas pipelines) and adding Detectron2 support to the Sahi toolkit, reflecting a blend of model research and pragmatic engineering. Now focused on omni-model architectures and training TTS models from scratch, he combines deep optimization expertise with dataset engineering to move research models into reliable production.
8 years of coding experience
2 years of employment as a software developer
Bachelor's degree Electrical and Electronics Engineering, Bachelor's degree Electrical and Electronics Engineering at Sivas Cumhuriyet Üniversitesi
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (16)
object-detection10
computer-vision10
pytorch10
machine-learning10
deeplearning-ai10
stable-diffusion10
deep-learning10
integrate10
detectron10
python10
integrations10
data-integration10
text2image10
tiling9
flax4
Programming languages (9)
Github contributions (5)
obss/sahi
Jan 2022 - Dec 2022
Framework agnostic sliced/tiled inference + interactive ui + error analysis plots
Role in this project:
ML Engineer
ML EngineerContributions:27 reviews, 7 commits, 15 PRs in 11 months
Contributions summary:Abdulkadir primarily focused on integrating Detectron2 support into the `sahi` framework. This involved modifying existing code and adding new functionalities to accommodate Detectron2 models for object detection and inference. The user made numerous updates to adapt the framework, including code refactoring, bug fixes, and the removal of legacy parameters, as well as improvements to exception handling. Overall, the contributions centered around enabling and enhancing the use of Detectron2 within the Sahi framework.
pythondetectron2tiledagnosticerror-analysis
huggingface/diffusers
Mar 2023 - Dec 2024
🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch and FLAX.
Role in this project:ML Engineer
ML EngineerContributions:1 review, 14 PRs, 76 comments in 1 year 9 months
Contributions summary:Abdulkadir made several contributions related to the `StableDiffusionTilingPipeline` and `StableDiffusionCanvasPipeline`, indicating a focus on developing and improving diffusion models. Their work involved supporting community-contributed pipelines, updating documentation, and refactoring code, which suggests involvement in expanding the library's capabilities and usability. Furthermore, the user integrated the FreeU mechanism, a key component in enhancing image generation models, which points to direct contributions to the core functionality of the diffusion models.
pytorchartdeep-learningimage2imagestate-of-the-art
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.