My name is Piotr Kluska. I am working in the APROPOS European joint Ph.D degree program at IBM Research Zürich and Universitat Politècnica de València since September 2021. I am originally from Poland, where I received my B.Sc. and M.Sc. in Computer Science from the Wroclaw University of Science and Technology in 2017 and 2019 respectively. During my studies, I have focused on the compression and quantization of deep learning models. Additionally, I have started working professionally in 2015 as an iOS Engineer with a change in career path to Machine Learning Engineer. I had an opportunity to work on multiple machine learning projects during my time at Outra, AXA XL, and intive. My research interests cover optimization of deep learning models especially focusing on computer vision, quantization, and programming languages. Outside of work, I enjoy programming, travelling, longboarding, and horseback riding.
The title of my research is “Efficient deep learning inference models for low-energy applications”. During my work on it, I will focus on the compression of deep learning models. Nowadays, deep neural networks have millions or even billion parameters that require abundant resources (memory, compute, energy) to process data. That makes those models hard to deploy at scale on edge devices. This project aims to introduce and evaluate novel algorithms for fast inference of the models on various edge devices, cloud, bare metal machines, and hybrid cloud. As a result, we should observe lower energy, memory, and resources consumption with as little as possible degradation of deep learning models performance.
Since my bachelor studies I have been fascinated with deep neural networks and challenges with deploying them to the edge devices. I am glad that I can participate and grow within APROPOS project as my thesis topic is aligned with my research interests. I hope that during this project I will be able to make contribution in deep learning optimization area and as a result grow as a researcher.
