Tech team profiles: Zarina Rakhimberdina

NOVEMBER 7th, 2024

BRAID’S TEAM

<aside> 💡 Welcome to another article in our series of Tech Team Profiles! Here, we’ll introduce you to the team building Braid’s automated engineering design technology. Zarina holds a PhD in Artificial Intelligence and has worked with fMRI brain data to extract information relevant to neurological disorders.

</aside>

Briefly describe your career thus far.

I completed my Bachelor's in Computer Science back in Kazakhstan. I was part of the first graduating class from my university, which was newly established. Today, the university ranks in the top 30% among international research universities worldwide. It was an amazing experience, not only because of the technical skills I gained, but also thanks to the interaction with outstanding professors invited from all over the world. They nurtured my interest in research activities.

After my Bachelor's, I decided to go to Japan for my Master’s, which wasn’t an obvious choice at the time; I had no family or connections here, but I wanted to experience something new. I went through a competitive selection process and was awarded the Japanese government’s MEXT scholarship.

I had the opportunity to explore both academia and industry through a series of interesting internships. I started as an intern for an online educational development program using machine learning to analyze course effectiveness in collaboration with Tokyo Tech (nowadays Institute of Science Tokyo), followed by a part-time position at a startup using LSTM models for electrical power prediction, and later completed an internship focused on MRI data analysis for diagnosing neurological conditions, which sparked my interest in graduate research.

As I moved from Master’s into my PhD studies, I decided to focus on publishing papers and completing research, so I didn't take on further internships. However, this period clarified that I enjoyed the hands-on nature of working with others on real-world projects and creating tangible products.

Once I finished my PhD, I joined a Japanese company that provides AI consulting services. My role mainly involved building computer vision models, such as a project for a warehouse environment to detect hazards. We trained vision models to recognize people, forklifts, and proximity levels between them to estimate potential danger zones.

Explain in lay terms your PhD thesis, “Improving automated detection of autism spectrum disorder with deep learning based on resting-state and task-based fMRI data.”

My PhD was interdisciplinary. I worked on medical data, specifically brain data from fMRI scans, to find meaningful patterns that could help neurologists. I used deep learning methods to analyze this data, which is quite complex because brain data can be represented in various forms. For example, you can look at it as a 3D volume or as temporal data (signals over time), or even as a graph showing how different brain regions communicate with each other.

I worked with both resting-state and task-based fMRI data. In resting-state fMRI, the brain's activity is recorded while the person is at rest, without performing any specific tasks. In task-based fMRI, the subject is engaged in a particular task, such as processing visual information, which helps us observe how specific regions of the brain respond. One of my projects involved attempting to reconstruct what a person was visually perceiving based solely on their brain activity. By comparing both resting-state and task-based fMRI data, I aimed to identify patterns that could help distinguish between typical and atypical brain function.

Zarina presenting at Women in Science Japan’s 2024 Flagship Event

Why did you decide to go into industry after completing your PhD, rather than continuing in academia?

During the final years of my PhD, I realized that the research experience wasn’t what I had expected. It's important to me that the results of my work have practical applications. Too often, experimental data and methods in the literature are not publically available, which can significantly slow down the research progress. While working on my thesis, I transitioned to the industry and found it rewarding, with greater opportunities for collaboration and real-world impact.

What originally sparked your interest in computer science?

I grew up in a technology-driven environment and saw many opportunities in computer science. My mother was a lecturer in information technology at a university. When I was little, the first IBM computers appeared in our city, and the demand for computer literacy sharply increased. She would take me with her to her evening computer literacy classes for middle school students. As a child, I knew the basic parts of the computer and the layout of the keyboard, and I would help students turn the computers on and off. Even though I was exposed to it at an early age, I didn’t seriously consider pursuing computer science as a career until after high school.

“Teacher assistant” after a rewarding day, ready to head home.

What’s a recent technological development that you’re excited about?

One of the things I’ve been really excited about is large language models (LLMs). I particularly like that there are not only closed-source models like OpenAI’s, but also open-source models that we can experiment with. For example, LLaMA 3 has a model with just 8 billion parameters, which you can run at home using a GPU. It’s amazing to see how far these models have come and how accessible they’ve become. However, it’s important to note that LLMs still have limitations, particularly when it comes to solving complex engineering problems, where they can’t outperform human level reasoning and critical thinking.

What’s the most interesting paper you’ve read recently?