Machine Learning in Healthcare - Insights from Our Senior ML Engineer : BlueGrid.io

In the rapidly evolving healthcare landscape, Stevan, our Senior Machine Learning Engineer, works to advance this field. With a strong background in computer science and bioinformatics, he uses cutting-edge technology to make significant improvements in medicine.

But what does he like the most about this role? How did he enter this field, and what drives his passion for it?
Find out directly from his experience.

What is your role at BlueGrid, and which industry do you work in?

I am a Senior Machine Learning Engineer working on a Computer Vision project in the field of digital pathology, part of the healthcare industry.

How did you get into this field, and what keeps you passionate about it?

While studying for my degree in Computer Science here in Belgrade some 15 years ago, I’ve also been a teaching associate at the Department of Physics at Petnica Science Center. I was mostly using programming to handle data analysis. Either to calculate a statistic, simulate a physics phenomenon, or visualize the data. There wasn’t a degree in Data Science or Machine Learning available back then. Most people we would call Data Scientists today have started with either programming, mathematics, or domain background. I was lucky enough to join a bioinformatics startup as a bioinformatician (a role that would be called Data Analyst today) right after graduating. During my career, I’ve worked with data from multiple domains. The idea of using cutting-edge technology to solve real-world problems, particularly those with the potential to improve lives, deeply motivates me. That’s why I’ve been working on multiple projects in the healthcare industry.

Can you describe a day in your ML role?

A typical day involves a blend of tasks, from refining machine learning models to analyzing datasets or experimentation results. Collaborating with domain experts from our team helps me better understand the intricacies of the pathology images we’re working with.

What tools and technologies do you use?

Regarding tools and technologies, we rely heavily on Python and libraries/frameworks like PyTorch and ClearML, for model development and evaluation, and OpenCV, Pillow, libvips, etc., for image manipulation. We also make use of cloud computing for scalability and efficient resource management.

What are the main goals and challenges you face as an MLE?

The main goals revolve around improving the accuracy and efficiency of our models. Challenges include dealing with noisy or limited data, addressing biases in the data, and ensuring the models are interpretable and reliable in a clinical setting.

Who do you work closely with on this project?

Collaboration is the key in this role. I’m very lucky to be a part of the great team assembled for this project. Next to MLEs and Data Engineers, we have wonderful medical doctors and their hospital staff working on getting us the best data available. There are also university professors and industry veterans with experience in digital image processing. All of them are there for us whenever we have a question about the data or domain in general.

From your experience, how do Data Science and Machine Learning intersect?

Data Science and Machine Learning intersect in various ways. 10 years ago there wasn’t much of a difference between those roles. All were under the umbrella of the Data Scientist, but we’re seeing a shift where there’s a clear distinction between Data Analysts, Data Scientists, Machine Learning Engineers, and Data Engineers. I should also mention the latest addition to that list, a role that would fit somewhere between DevOps and Machine Learning Engineer: MLOps Engineer. Fundamentally, Data Science provides the tools and techniques for extracting insights and patterns from data. Machine Learning leverages those insights to build predictive models and make decisions.

How important is communication in the success of your ML activities?

Effective communication is crucial for the success of any Machine Learning project across various stages of development and deployment. It ensures that engineers understand project needs, and collaborate effectively with Data Scientists and domain experts for proper data preparation and model selection. Communication also facilitates the seamless integration of ML models into existing systems. Addresses feedback for iterative improvements, and ensures compliance with ethical and legal considerations has been a growing problem in the industry for quite some time.

What do you like the most about this role?

It is the opportunity to make a tangible impact on healthcare outcomes. There’s nothing (work-related) like knowing you’ve made a difference in people’s lives, be it through speeding up a medical procedure, making it more accurate, or a combination of both. Solving interesting engineering problems all the while comes in a very close second, though.

What technical and other skills do you find most valuable in your role?

In terms of technical skills, proficiency in Machine Learning and, to a lesser extent, Software Engineering is essential. Additionally, qualities like adaptability, problem-solving ability, and a genuine curiosity about the domain are invaluable.

How do you stay updated on trends in Machine Learning and Healthcare?

I find inspiration and stay updated on trends through various channels. Some of them are research papers, networking with peers in the field, and following their posts on social networks. Keeping an eye on advancements in both Machine Learning and healthcare and biotech domains helps me a lot. However, it cannot be stated enough that doing that thoroughly is a full-time job in and of itself. Some of the resources I can wholeheartedly recommend are:

3Blue1Brown’s YouTube playlist on Neural Networks: https://www.3blue1brown.com/topics/neural-networks,
Following Sebastian Raschka on X and reading his blogs: https://x.com/rasbt and https://magazine.sebastianraschka.com/archive,
Subscribing to The Batch newsletter: https://www.deeplearning.ai/the-batch/.

My advice would be to continue honing your technical skills. Also, cultivate a genuine passion for leveraging technology in solving real-world problems. Be open to learning from others, stay curious, and never underestimate the power of effective communication in driving project success.