GSoC 2026 - Optimize Quantized Model Inference Performance on ARM Devices with OpenVINO

[selcuksntrk]

Hi @alvoron @v-Golubev ,

I'm Selcuk Senturk, an AI/ML Engineer at IBM Expert Labs and MSc Computer Science student at Ege University. I'm interested in contributing to this project as part of GSoC 2026.

My background on the ML side is directly relevant, I work daily with PyTorch-based model training, optimization, and production deployment at IBM, including model deployment on NVIDIA GPUs with inference optimization, working with LLM deployments where latency, throughput, and memory footprint are real production constraints. This gives me concrete intuition for what quantization schemes are actually trading off at runtime.

My BSc in Physics gave me strong foundations in numerical methods and linear algebra, which I find directly applicable to understanding what quantization schemes are actually doing mathematically and where performance is lost.

On the systems side, I know C++ syntax and core concepts like templates, memory model, RAII, low-level data structures and I've read through C++ codebases including parts of inference runtimes. I don't have professional C++ experience, but I'm committed to ramping up quickly and I'm comfortable working at that level.

On hardware, I have an M-series Mac (ARM) as my primary machine and access to other ARM-based devices, so the hard requirement there is covered.

The NEON vectorization and ARM Compute Library integration angle is what I find most technically interesting, the intersection of numerical optimization and hardware-aware programming.

Also I want to note that I don't want to limit my contributions to the GSoC timeline, I would like to continue to work on this topic.

I would like to discuss further if I'm a good candidate and how I plan to implement this feature. I really want to work on this project since I want to go deeper in ml systems engineering in my career.

Thank you, have a good day.

Resume: Drive
LinkedIn: selcuksntrk

[alvoron]

@selcuksntrk
I've prepared the instructions here: https://github.com/alvoron/gsoc-2026-openvino/tree/main
Please don't hesitate to ask if you have any questions!

[selcuksntrk]

@alvoron

Thank you very much, I'm very excited to apply this project. I am going to solve the test assignment as soon as possible before submitting the proposal. Have a good day.

[alvoron]

I added a new section to README, please take a look: https://github.com/alvoron/gsoc-2026-openvino/tree/main?tab=readme-ov-file#6-start-addressing-a-technical-gap-optional

[selcuksntrk]

Thank you for the update @alvoron . I am reviewing the previous commits and PRs you added to the README to fully understand them. Although it is sometimes difficult to understand certain library functions, the commits are very helpful for understanding the quantization implementation in the OpenVINO repo. Also, I’ll definitely review the new update. I have started writing the proposal, but I would like to have your feedback before submitting it to the GSoC website.