Career Nav #76: My Unique Path to Data Science

Career Nav #76: My Unique Path to Data Science

Written by Shannon Stewart


iTunesSpotifyGoogleVideoMore Episodes

Shannon Stewart, Principal Research Scientist at Altana AI, shares her talk, “My Unique Path into Data Science.” She discusses her path from Molecular Biology and Chemical Senses to Data Science, how she ended up in Taiwan and Holland, and all the steps in between that have shaped her career journey.

My education was in Molecular Biology, and my PhD was in Chemical Senses. You might be thinking, you must have made a really big pivot to get into data science, but I don’t see it that way. The data that I worked with as a PhD student was multi-faceted. It was chaotic because some of it was animal behavior data. This shares some important features with the kinds of data humans and human institutions generate.

The techniques that I learned to analyze that kind of data are things I still use in my career today. I also use the path of thinking that I learned as a scientist to design good experiments that get us to the root cause of what we observe in data. My first detour happened right away, I did not have a job lined up after my PhD. I got an opportunity to go to Taiwan, and I took it.

While in Taiwan, I took Intensive Language classes, which were pivotal and important to the rest of my journey. It introduced me to a new kind of flexibility of thinking and empowered me with many new skills that I hadn’t had before, not just language skills but resiliency. It was also a radical step, forcing me to get far outside my comfort zone and trust that I could muddle through discomfort.

I had gone from the top of my field, finishing a PhD, to being completely illiterate. It was a huge challenge. I still use my skills to get there in my work today. My first professional job was at the MIT Center for Biomedical Innovation. I was working on a project to identify the leading indicators that food or medicine had been adulterated on purpose so that the FDA could stop those shipments from getting into the country. Here’s where I started learning about modeling supply chains and using AI techniques.

Our team was highly interdisciplinary; we had members from MIT’s Sloan School of Business, CSAIL, and the computer science and AI laboratory. I got to learn from every member of the team, even students. One thing that I picked up from them was how to start a coding project on a completely new problem where no solution path exists. I also learned a ton about Natural Language Processing, or NLP. Another key piece of knowledge I picked up was how to tie observations or features in the data back to their actual root causes, which is an important skill in data science. You have to do this to avoid, for instance, picking multiple features correlated to the same cause. You want your models to be robust and reflect what you’re trying to measure or predict. 

MIT is also synonymous with startups and innovation. I took a class at the Sloan School of Business on doing a start-up, and it was way more accessible than I thought. I left that class with the idea that I wanted to try it out. The other thing that I started thinking about at this point was innovation, not just a buzzword, but how to identify these situations where big systems conflict with each other and they’re going to create an opportunity that has not existed before.

Conversely, how to intentionally break a system that has existed for a long time, but it shouldn’t, or it’s not working. One thing that came up toward the end of my time at MIT was when I traced back why we saw so many shipments rejected for a certain food product. Once I traced the supply chain, I learned more about the supplier.

There were local news reports in the native language of its context that showed that they had workers living in iron cages on the premises. It was a huge local scandal. It opened my eyes to the fact that literal slavery is still happening in the world today. To this day, this is one of the most egregious examples of this I have ever seen. And since then, I have worked for years on forced labor. 

Meanwhile, I thought I had my next job in the bag. About a week before I started, it fell through catastrophically. When my husband got a job at a Dutch University, I left and moved to Holland. I felt completely defeated. It was an awful experience, and it undermined my confidence. When I tell the polished version of this story, I don’t mention this, but these setbacks are a key formative piece of the story as a whole.

Once I was established in Holland, a professor reached out to me who had come into possession of some business data showing North Korea was sending workers to Europe. The state was keeping most of those workers’ wages. This practice was financing their nuclear weapons program and buying luxury goods for their top elites. This practice happens in a shocking number of countries around the world. North Korean workers are building battleships, working in garment factories, doing all sorts of skilled and unskilled work, and they’re not receiving their wages.

They’re under a very strict system of control. I structured that data into a graph database based on how the ICIJ did it for the offshore leaks database. If you’re unfamiliar, that’s where the Panama and Paradise Papers are accessible. To do this, I had to teach myself, sign for query language, which I do not recommend. If you need to build a graph database, try to find a class. The documentation was very bad, but the theme here is I will overcome anything to teach myself to do anything if the goal is worth it. 

I was hired at The Global Fund to end modern slavery. It was like a start-up. I was the 14th employee. There were opportunities to define your role in the organization and try things out. I think the most important parts for me were clarifying my big vision and personal mission: to end modern slavery by making it unprofitable and smashing the business model. That’s where I started to see a path. I integrated what I learned at MIT about innovation and what I had learned from my practice about how forced labor exists worldwide.

My keystone project at defense was developing a machine learning classifier that takes in information about a company’s operations, which includes things like its location, financial health, trade patterns, and ownership structure, and it estimates based on its similarities or differences with firms that are known to use force labor, whether an unknown, unaudited firm has a potential risk and needs to be audited.

I’m a research scientist at Altana AI, where all these pieces occur. Altana was founded to pioneer a shared source of truth and empower governments, logistics providers, and businesses to build trusted, sustainable, ethical supply chain networks. Every experience I’ve had up to this point leads me to believe that this is possible and that our approach is an important first step. I was able to get repeatable signals out of the same data for adulterated and counterfeit pharmaceuticals as well as forced labor. At Altana, our knowledge graph includes over 400 million companies and billions of transactions, so we’re talking big data.

This job has been a great learning experience for me. For the first time, I’m working with really experienced software engineers. I’m learning so much from them, but we work together as a team using our unique strengths. Day-to-day, I mainly work on our core back-end technology, continuously constructing the knowledge graph. Within the team that does that, I lead the overlay of data sets that include things like lists of sanctioned companies or, conversely, lists of companies with important certifications. This type of task is one of the core functions of data science.

Reconciling a list on one hand with a list on the other hand, where the entities and records may or may not match across those lists. I also sometimes get to work directly with clients, helping them understand where their value chains intersect with risks like forced labor so that they can make informed decisions about whether remediation is possible or if they need to redirect their sourcing.

I also take on special projects that help our product work better, making the graph connect as it should by enriching it if needed. Sometimes, this uses skills I learned deep in the past, like the Chinese language NLP. It is so rare to have a path with no setbacks and diversions, and no amount of planning and preparation will help you avoid them. Everybody has them. In my case, the year I spent in Taiwan, I didn’t know it would lead to anything productive for me, but I had always wanted to live overseas. I regretted not participating in something like studying abroad.

When I got that chance, I just took it. Following my path and being authentic opened up something that hadn’t been possible before. I didn’t have any particular goal in mind when I did it. I just could not have imagined what the future held for me. That’s what my unique path taught me: trusting yourself and being true to yourself will bring the goal into focus. Being pragmatic in planning doesn’t necessarily produce better or more predictable success. You’ll have setbacks. They might be big, but they’re also an opportunity to refocus on what’s important to you and become your authentic self, someone you might not meet without those challenges. 



Guest: Shannon Stewart, Principal Research Scientist, Altana AI 

Producer: JL Lewitin, Senior Producer Press and Communications, Women Who Code