Life at Nautilus interview with Winnie Leung – Software Engineer

Biology researchers capture fascinating images of life taking shape and taking action. From photos of condensed chromosomes getting ready for cell division to videos of wild dogs hunting in the grasslands and forests of Africa, there’s plenty to inspire awe.

Winnie Leung, Nautilus Software Engineer, spent her masters capturing similarly awe-inspiring videos of isolated muscle cells beating in petri dishes. She would coax induced pluripotent stem cells to become these “cardiomyocytes” and, while she was drawn to how cool stem cells are, she knew she needed to measure their properties to understand if they really did differentiate and mature into functional muscle cells. To make her observations quantitative, she automated the analysis of stem cell images and defined parameters indicative of cardiomyocyte functionality. Thus began Winnie’s career developing sophisticated image analysis skills that she expertly puts to use at Nautilus today.

What do you do at Nautilus?

At Nautilus, I help develop the software tools used to analyze images taken by the Nautilus Platform. As we measure proteins and proteoforms, we generate many, many images. These show whether our probes bind to proteins deposited on the platform’s flow cells and are the raw material for protein quantification. With images containing billions of single protein molecules, it would be impossible to analyze everything by hand. My team develops the robust image analysis pipelines for measuring and classifying probe binding events. These pipelines are essential for decoding proteins and identifying proteoforms.

What do you enjoy about your image analysis work?

I love that I’m given the freedom to creatively solve problems by exploring a wide variety of solutions. My co-workers and I are particularly motivated when we get to work on a novel challenge and learn something new, so we thrive on problems that require us to dig into the latest software engineering, image processing, or data analysis techniques.

For example, with the huge number of images we generate, some portion of them might be of “bad” quality or have regions that are out of focus. We’ve built an image simulation tool to help quantify how “good” certain image properties should be for us to confidently extract data from them. These properties were used in turn to define standards of quality with quantifiable metrics shared across the organization. Extending the usefulness of this work, we automated the identification of “good” image regions that are suitable for downstream analysis. This enables us to retain more useful data from our images. When you look at an image, your brain can quickly determine whether it’s “good” or “bad” but defining “good” or “bad” for our software pipeline was a fascinating challenge. Putting our solutions into practice has been very satisfying.

I also love that working in image processing enables me to interact with many other teams and work on cross-team projects. For example, after defining the metrics we extract from our images, I was surprised to learn how many people were interested in analyzing them. To make this useful data available to others, I worked cross-functionally with another team to build a pipeline that generates and stores all our metrics on AWS. This makes the metrics easy to access on the cloud and enables everyone to do further analysis and validation. It was hugely gratifying to work with other teams and create something that extends the impact of everyone’s experiments.

If you have a difficult day what keeps you motivated?

Knowing that my team and I are working towards a solution. Even if the solution to a problem is not clear, and I’m a bit frustrated, I love knowing that everyone around me is super competent and that we’re pooling our efforts toward a common goal. It’s great to have team members who are genuinely interested in helping one another and who will always join a quick brainstorming session anytime we need to work through a problem. The camaraderie is fantastic.

Even with this great teamwork, I can get frustrated by processes that I must repeat manually over and over. Thankfully, a big part of my job involves developing means to automate such processes. Knowing I have the skills required to automate tedious tasks puts me at ease.

What path do you hope your career will take moving forward?

I strive to solve hard problems. For me, a hard problem is one that forces me to learn something new. As someone who tries to stay at the cutting edge of software engineering, I think my next big challenge will take the form of developing and applying my understanding of machine learning. Once we start churning out large amounts of proteomic data from the Nautilus Platform, I hope to help integrate some new machine learning concepts into our image processing pipelines. This seems to be the direction that a lot of biological analysis is moving, and I’m excited to be a part of it.

What excites you most about Nautilus’ future?

When I interviewed at Nautilus, I was struck by how important it is to develop a technology that can quickly analyze proteins – the actual molecules that do most of the work in cells. Back in my masters, measuring proteomes would have been incredibly useful in determining how well our stem cell differentiation and maturation protocols worked. I think most biologists could similarly put proteomics to great use if it were easy. I can’t wait to give researchers the ability to accessibly investigate the full proteome.

On a more personal level, I have family members with Alzheimer’s and I’m very excited about the work we’re doing on tau proteoforms. Being able to understand how the tau protein differs between people with and without Alzheimer’s could accelerate the creation of biomarkers and drugs for the disease. It’s fantastic that we can already generate novel tau proteoform data, and I’m hopeful that data can be applied to help people soon.