Sylwia Majchrowska

Sylwia Majchrowska

Postdoc Research Fellow

AstraZeneca

About me

Hi, I am a passionate AI enthusiast and problem solver, currently postdoc fellow in computer vision domain. With a background in physics and biomathematics, I joined the esteemed Eye for AI program led by AI Sweden, where I explored the practical applications of AI in domains such as healthcare, autonomous driving, and pharmacy. Through my research experiences, I gained valuable insights into data-driven approaches and honed my technical skills in Python, data science, and machine learning. I actively contributed to ongoing projects, collaborating with experts in the field and developing essential professional skills in communication, teamwork, and project management. Now, I am eager to share my knowledge, open to new collaborations, and passionate about leveraging AI to solve real-world problems.

Interests

  • Computer Vision
  • Data-driven approches
  • Mathematical modelling
  • Unsupervised learning

Education

  • PhD in Physics, 2017 - 2022

    Wroclaw University of Science and Technology

  • MSc in Mathematics, 2016 - 2017

    University of Wroclaw

  • MEng in Technical Physics, 2016 - 2017

    Wroclaw University of Science and Technology

  • BEng in Technical Physics, 2012 - 2016

    Wroclaw University of Science and Technology

  • BS in Mathematics, 2012 - 2016

    University of Wroclaw

Skills

Python

Nonlinear simulations

Teaching

Professional Experience

 
 
 
 
 

Postdoc research fellow

AstraZeneca

Jun 2023 – Present Gothenburg
Harnessing the potential of AI to overcome data limitations in the medical domain.
 
 
 
 
 

Data Scientist

AI Sweden

Oct 2021 – Apr 2023 Gothenburg
As a participant of the exclusive Eye for AI program, developed by AI Sweden in collaboration with AstraZeneca (pharmacy), Sahlgrenska University Hospital (healthcare), and Zenseact (automotiv), with support from Women in AI, I worked on tangible applied data and AI challenges across three diverse application domains.
 
 
 
 
 

Deep Learning Researcher

NeuroSYS

Jul 2019 – Oct 2021 Wrocław
Research on computer vision, signal processing, data cloud processing, and deep learning.
 
 
 
 
 

Junior Data Analyst (trainee)

Nokia

Oct 2018 – Jun 2019 Wrocław
Automation with scripting to support data organization and decisions by creation of user-friendly tools visualizing the financial domain of the benchmarked product or service.
 
 
 
 
 

Numerical Simulation Researcher

XTPL

Jul 2017 – Sep 2018 Wrocław
Developing a numerical model describing the physical phenomena occurring during the formation of ultrathin conductive lines.

Recent Publications

Quickly discover relevant content by filtering publications.

Intermodal-vectorial four-wave mixing processes involving LP01, LP11, LP02 and LP21 modes of birefringent fibers

We present the complete (analytical, numerical and experimental) analysis of intermodal-vectorial four-wave mixings proccesses in birefringent fibers. We analyze phase-matching condition and overlap coefficients to indicate possible processes. Then, we demonstrate multiple four-wave mixing processes in LP01 and LP11 modes numerically and experimentally. Finally, we extend theoretical analysis to account higher-order modes, LP02 and LP21.
Karol Tarnowski, Kinga Żołnacz, Sylwia Majchrowska, Wacław Urbańczyk
PDF DOI
Intermodal-vectorial four-wave mixing processes involving LP01, LP11, LP02 and LP21 modes of birefringent fibers

Unlocking the Heart Using Adaptive Locked Agnostic Networks

We introduce the Adaptive Locked Agnostic Network (ALAN), a concept involving self-supervised visual feature extraction using a large backbone model to produce anatomically robust semantic self-segmentation. In the ALAN methodology, this self-supervised training occurs only once on a large and diverse dataset. We applied the ALAN approach to three publicly available echocardiography datasets and designed two downstream models, one for segmenting a target anatomical region, and a second for echocardiogram view classification.
Sylwia Majchrowska, Anders Hildeman, Philip Teare, Tom Diethe
PDF Code DOI
Unlocking the Heart Using Adaptive Locked Agnostic Networks

Assessing GAN-Based Generative Modeling on Skin Lesions Images

We evaluated models’ performance in terms of fidelity, diversity, speed of training, and predictive ability of classifiers trained on the generated synthetic data. In addition we provided explainability through exploration of latent space and embeddings projection focused both on global and local explanations.
Sandra Carrasco Limeros, Sylwia Majchrowska, Mohamad Khir Zoubi, Anna Rosén, Juulia Suvilehto, Lisa Sjöblom, Magnus Kjellberg
PDF DOI
Assessing GAN-Based Generative Modeling on Skin Lesions Images
See all publications

Recent & Upcoming Talks

Leveraging Synthetic Data for Skin Lesion Analysis

This research delves into the application of unconditional and conditional Generative Adversarial Networks (GANs) in both centralized and decentralized settings. The centralized approach replicates studies on a large but imbalanced skin lesion dataset, while the decentralized approach emulates a more realistic hospital scenario with data from three institutions. We meticulously assess the models’ performance in terms of fidelity, diversity, training speed, and the predictive capabilities of classifiers trained on synthetic data. Moreover, we delve into the explainability of the models, focusing on both global and local features. Crucially, we validate the authenticity of the generated samples by calculating the distance between real images and their respective projections in the latent space, addressing a key concern in such applications.
Oct 3, 2023 2:00 PM — 5:30 PM Paris Convention Centre
Sandra Carrasco, Sylwia Majchrowska
Code
Leveraging Synthetic Data for Skin Lesion Analysis

Unlocking the Heart Using Adaptive Locked Agnostic Networks

We introduce the Adaptive Locked Agnostic Network (ALAN), a concept involving self-supervised visual feature extraction using a large backbone model to produce anatomically robust semantic self-segmentation. In the ALAN methodology, this self-supervised training occurs only once on a large and diverse dataset. We applied the ALAN approach to three publicly available echocardiography datasets and designed two downstream models, one for segmenting a target anatomical region, and a second for echocardiogram view classification.
Oct 3, 2023 2:00 PM — 5:30 PM Paris Convention Centre
Sylwia Majchrowska, Anders Hildeman, Philip Teare
Code
Unlocking the Heart Using Adaptive Locked Agnostic Networks

Overcoming Barriers to Data-Driven Workflows in Healthcare and Mobility

This work explores a usability of synthetic data for training visual deep learning models in real-world applications.
Apr 25, 2023 8:00 AM — 12:00 PM Ericsson office, Gothenburg
Sylwia Majchrowska, Sandra Carrasco
Slides
Overcoming Barriers to Data-Driven Workflows in Healthcare and Mobility
See all talks

Research

*
All Physics Nonlinear optics Deep Learning Computer Vision Healthcare

Interaction prediction using GNNs

The main area of interest of the project is research on explainable and interpretable motion prediction approaches of road users in highly interactive scenarios using NuScenes dataset.

Decentralized AI in Healthcare

Create shared models together with both, or several, regions, and discover how Swedish hospitals can collaborate in a practical sense.

Synthetic data generation

Using generative modelling techniques for generating synthetic data of skin diseases.

Hear AI

Non-profit project with the aim to use AI for Polish Sign Language.

Modelling materials with complex electromagnetic properties

gprMax is open source software that simulates electromagnetic wave propagation. It uses Yee’s algorithm to solve Maxwell’s equations in 3D using the Finite-Difference Time-Domain (FDTD) method.

Automatic documents, images and video processing

Developing deep learning algorithms for speech enhancement, optical character recognition, and document segmentation.

Augmented reality for industrial processes automation

Developing deep learning algorithms for speech enhancement, keyword spotting and gesture recognition.

Detect waste in Pomerania

Using detection models to localize and classify waste on images and video.

Nonlinear phenomena in multimode fibers

Modeling of nonlinear propagation in multimode fibers - influence of fiber parameter on nonlinear phenomena. Investigation of multimode solitons and frequency conversion.

AI-generated company visual identity

Business needs analysis, a feasibility study, and developing a system to generate a new logo based on images uploaded by the community using deep learning.

Recognition and identification of bacterial colonies

The aim of the project is to develop a method to automate the analysis of bacterial colonies on Petri dishes using artificial neural networks and Machine Learning algorithms.

Nonlinear light propagation in multimode fibers

Project aims to develop more accurate numerical tools to model nonlinear propagation in multimode fibers and to investigate the control of nonlinear interactions with new fiber design.

A new generation of TCF layers for use in displays and thin film photovoltaic cells

The subject of the project was the development of technology for the production of a new generation of transparent conductive layers (TCF).

Communities

*
All Professional Experience Education Volunteering

AstraZeneca

AstraZeneca is focused on creating genuinely innovative medicines and improving access to them. In this way, the company delivers the greatest benefit to patients, healthcare systems and societies globally. The organisation has three priorities designed to deliver the growth through innovation strategy - accelerate innovative science, deliver growth and therapy area leadership, and be a great place to work.

EU TalentOn

Between September 14 and 18 2022, Leiden European City of Science 2022 was filled with Bright Young Minds. Among them, a 104 participants to the EU TalentOn and me as one of them. United in teams, we were challenged to come up with the best scientific solutions to societal challenges, such as, Climate change, Cancer, Restoring waters, Smart cities, and Healthy soils. All the while expanding knowledge and connecting to industry and institute partners.

Zenseact

Moving towards Zero collisions faster with the help of autonomous driving technologies. We are passionate about life and safety. Our goal is to protect life on the road by providing a software platform for innovative and safe self-driving features that will change our societies forever. The development of autonomous driving has proven to be the engineering challenge of the century. Solving it requires razor-sharp focus and that’s why Zenseact exists.

Sahlgrenska University Hospital

Sahlgrenska University Hospital is one of Sweden’s largest hospitals - improves healthcare through the application of strong innovation and clinic research in close and broad cooperation with academia, industry and patients.

AI Sweden

The Swedish National Center for applied AI, supported by the Swedish government and the public and private sectors across the country. The mission is to accelerate the use of AI for the benefit of our society, our competitiveness, and for everyone living in Sweden.

Women in AI Poland

Women in AI (WAI) is a global nonprofit do-tank on a mission to increase female representation and participation in AI. They do that by organizing educational events and programs, working with both corporate and startup ecosystems to build an unbiased and fair future.

gprMax

gprMax is open source software that simulates electromagnetic wave propagation. It uses Yee’s algorithm to solve Maxwell’s equations in 3D using the Finite-Difference Time-Domain (FDTD) method. The finite difference expressions for the spatial and temporal derivatives are central-difference in nature and second-order accurate.

Women in Machine Learning & Data Science Foundation

WiMLDS’s mission is to support and promote women and gender minorities who are practicing, studying or are interested in the fields of machine learning and data science. We create opportunities for members to engage in technical and professional conversations in a positive, supportive environment by hosting talks by women and gender minority individuals working in data science or machine learning, technical workshops, networking events and hackathons. We are inclusive to anyone who supports our cause regardless of gender identity or technical background. However, in support of our mission, priority for certain events and opportunities will be given to women and gender minority members.

CoderDojo Foundation

The CoderDojo Foundation has vision - every child worldwide should have the opportunity to learn code and to be creative with technology in a safe amd social environment.

Girls Code Fun Foundation

Girls Code Fun is a foundation whose aim is to encourage more school aged girls in Poland to pursue an education and/or career in technology. Our goal is to show girls that programming and technology is something for them. We offer programming courses for children ages 5-19 and one-time programming workshops for both children and adults!

Fiber Optics Group

The Fiber Optics Group possesses an expertise and necessary instrumentation for investigation of conventional and specialty fibers (including photonic crystal fibers) and a variety of fiber-optic components and devices.

NeuroSYS

Technologies disrupt businesses and whole industries. Willing to be a part of that change, we work on introducing benefits of augmented reality, artificial intelligence, and deep learning to our daily life.

Perspektywy Education Foundation

Perspektywy Education Foundation (Fundacja Edukacyjna “Perspektywy”- in Polish) is an independent, non-profit national organization established June 1st, 1998 to promote and support education. I take part in IT for She project, which is a volunteering program which will send active young women studying IT to small towns in summer, to teach programming and tech knowledge at schools.

PROJECTOR Student Volunteering

PROJECTOR Student Volunteering is a program implemented by the Educational Enterprise Foundation. The aim of the Program is to activate and counteract the exclusion of children and young people from small towns and villages all over Poland by developing their passions and interests, discovering talents and potential hidden in both students and project participants.

NOKIA

Driving innovation for tomorrow and delivering technology today, we make businesses more productive, environments cleaner, workplaces safer, economies stronger and people’s lives richer.

Laboratoire Interdisciplinaire Carnot de Bourgogne

The Laboratory Interdisciplinaire Carnot de Bourgogne (ICB), jointly operated by CNRS, University of Burgundy and University of Technology Belfort-Montbéliard, counts 300 Physicists, Chemists, Engineers and Technicians in Dijon, Le Creusot, Châlon-sur-Saône & Belfort (Sevenans) campuses. They develop new optical functionalities and new materials for industry, medicine and telecommunications.

XTPL

XTPL is a company developing globally innovative, additive manufacturing technology that enables ultra-precise printing of nanomaterials. Unique XTPL printing system allows for precise deposition of an in-house formulated nanoink on a variety of substrates to obtain conductive and nonconduvtive submicron structures.

University of Wroclaw

Our university was founded in 1702 by Emperor Leopold I Habsburg. It was named the Leopoldina in his honour. The first academic year began on 15 November, on the Emperor’s nameday.

Wroclaw University of Science and Technology

Wrocław University of Science and Technology was established in 1945, mainly as a result of the involvement of the academic staff of the now-defunct Technical University of Lviv and the Jan Kazimierz University in Lviv, who adapted the destroyed buildings of the German School of Technology - Technische Hochschule.

Contact