Tessa Xie reflects on the twists and turns of data science careers, the benefits of consulting, and the value of writing for a wide audience.
In the Author Spotlight series, TDS Editors chat with members of our community about their career path in data science, their writing, and their sources of inspiration. Today, we’re thrilled to share our conversation with Tessa Xie.
Tessa started her career in finance as a quantitative researcher after graduating from MIT with a degree in financial engineering. She later worked in data science consulting at McKinsey, followed by a transition into the tech industry as a data science manager at Cruise, an autonomous driving company. Most recently, she started her new journey as a data science manager at LinkedIn.
How did you decide to become a data scientist?
I first became interested in data science in graduate school, where I majored in financial mathematics. Because there is a lot of overlap between financial mathematics and data science, we had the chance to take data science classes as electives for the curriculum. To my surprise, I found that I was more interested in the data science classes than the ones that focused specifically on finance.
However, I didn’t immediately enter data science after graduating, because I wanted to try out the industry I originally set out to explore — finance. After being a quantitative analyst for over a year, my initial hypothesis was confirmed: I like working with data, just not necessarily finance data.
I knew I wanted to leave finance but was not sure which industry I wanted to enter. Consulting seemed to be the best way to explore data science in different industries before settling into one. Consulting firms also interview data scientists from more diverse backgrounds than, say, tech, so it was a good opportunity to transition from finance to data science. Even though there’s overlap between quantitative finance and data science, I still decided to take a couple of online classes to refresh my data science knowledge, and those definitely paid off in interviews.
You recently spent some time in the autonomous vehicle sector. What is it like to work as a data scientist in that industry?
Because autonomous driving is such a futuristic field, one of the biggest challenges is the fact that we are solving a lot of data science problems for the first time. It’s a challenge because when you are thrown into a problem, more often than not you don’t have playbooks to learn from; you don’t have best practices that you can look up. It’s also an opportunity, because this means that as a data scientist you can take a lot of initiative, and it’s a great learning experience—not to mention that you will be able to drive meaningful impact even as a relatively junior data scientist.
Another challenge that I have noticed is that there is a huge amount of data being collected across different channels (e.g. lidar, radar, camera, etc.) for traffic or road-related information. Being able to efficiently and accurately connect data from those disjointed channels is crucial, since this will serve as the foundation of any data science analyses built on top. It provides an opportunity for data scientists to work super closely with data engineers and data infra teams. In the process, I was able to learn about data engineering and data infrastructure and appreciate the complexity involved in these areas.
The third point that’s worth mentioning is that since most autonomous driving companies are either still in the research and development stage or just started to commercialize their product, there’s not much customer data that you can work with as a data scientist. So if you are a data scientist who’s very interested in product analytics, user research, and/or experimentation, just know that you won’t have a huge amount of data to drive insights from (yet).
Do you have any other pointers for data scientists who are considering this industry?
When it comes to evaluating opportunities, I currently put two factors above others — learning opportunities and my interest and passion for the product. I was initially attracted to the autonomous driving field for exactly these two reasons. Because of the fact that it’s such a novel field, I knew there would be a lot of challenges when it comes to data. In my opinion, most challenges are opportunities in disguise; it’s the best way to force you to step out of your comfort zone and learn about new things. So I was very excited to enter a field that nobody has fully figured out and participate in the “zero to one” transformation.
Also, autonomous driving vehicles are a product that I have faith in and am passionate about. As a person who doesn’t know how to drive, I’m very excited about the possibility of autonomous vehicles and their potential. Not to mention that AV technology will hopefully be safer than human drivers and potentially save tons of lives we lose in car accidents every year. Being able to work on a technology that brings positive impact to the world and that I will be using one day made the work that much more interesting and meaningful.
You recently wrote about the experience of leaving a previous role as a McKinsey consultant. What questions should data scientists ask when they make a decision about joining or leaving a company?
Data science is a broad term that encompasses many different roles across a large number of industries, and it can be difficult for outsiders to get a good overview. To help with this, I have written a couple of articles about how to decide what type of companies you want to join and how to decide on the right role in the data world.
When deciding between companies, the biggest distinguishing factor in my experience is the size of the company. Joining a big corporation usually means you have more best practices to learn from and better, or at least more predictable, compensation, but at the same time less direct impact and potentially a more process-intensive corporate culture.
Smaller companies, on the other hand, usually provide more opportunities to take initiative and drive impact because a lot of things are not quite figured out yet. Contrary to the corporate culture, you will experience more of a hustle culture in smaller companies because things need to move fast. I went into detail about this in my article about whether you should choose to join a small startup or a big corporation. Of course there are outliers to any generalization, so you really need to judge companies case by case; but the type of questions you should ask yourself when making this decision should be more or less around the same dimensions — company culture, learning by doing or learning from best practices, compensation, and career outlook.
What about deciding which specific role to pursue?
When it comes to the type of role you want to choose, it’s a harder question because there are so many roles that can fall under the data science umbrella. There are data engineers, data scientists, data analysts, decision scientists, ML engineers, and more; it doesn’t help that a lot of companies use those titles interchangeably. But the most important question to ask yourself is: What do you like more, to explore problems or to implement solutions?
The more engineering components the role has (MLE, data engineer), the more it will be focusing on the “how” and you will spend most of your time on implementing solutions. The roles that are closer to business teams (data analyst, data scientist) will be focusing more on exploring the “what” and the “why.” As long as you know the answer to this, you should have an easier time mapping the job description to your interests.
Should these questions change depending on the career stage you’re at?
If you are in an entry-level position, these questions are less important—for the early years of your career, the most important thing is to learn. So the only thing that you really need to ask yourself is, “does this role provide me enough learning opportunities?” You will have plenty of opportunities to pivot your career if this role doesn’t fit all of your criteria. If you are in your mid-career, you will have an easier time answering the questions above based on your years of exploring and learning, and it becomes more important to focus on developing a clear profile that fits the career path you are interested in, as hiring managers and recruiters have increased the background requirements for experienced candidates.
You’ve written extensively about the interviewing experience for data science roles — why is it important for you to share these insights with a broader audience?
When I first decided to transition into data science, there were not as many resources online about how to prepare for DS interviews as there were for other roles, especially for people looking to switch from other types of roles and industries, such as finance. There’s the famous Leetcode for software engineers, and a book, Crack the PM Interview, for product managers. But when it comes to data science, I was more or less on my own for interview prep because it was such a new field.
After several rounds of recruiting, I have interviewed with countless companies, ranging from big tech to small startups and consulting firms. During this process, I have gotten a lot of insights from friends who are in this field, and have done a lot of research myself. After my job switch, I was getting a lot of outreach from aspiring data scientists through LinkedIn asking for my advice on interview prep. I’m big on supporting peers and avoiding repetitive work, so I thought it would be helpful for me to share my own interviewing experiences in a scalable way with a broader audience on a public platform like Medium.
I have recently started a new series specifically for interview topics covered in typical DS interviews. The goal is for readers to be able to use this series as a centralized starting point for their interview prep and job search. Speaking of which, I would love to get feedback from my audience on what would be helpful for them in their job search; I will try my best to incorporate those requests in future posts.
Finally, as a prolific author, do you have any advice for data professionals who might be interested in writing for a wide audience?
It’s a common thing for aspiring writers to suffer from imposter syndrome. I was definitely thinking, “I’m not a VP of Data Science nor am I a PhD in machine learning, how can I possibly add value to the data community?”
My advice is to forget about what you are not, and focus on what you are. It doesn’t matter who you are and what your experience is, you have something to bring to the table, and you can potentially help others with those insights. So finding that niche is key. If most of your work as a data professional is building models, I’m sure you have a lot of best practices you can share with people who are just starting out with modeling.
When it comes to finding new ideas for articles, I have three suggestions that I found helpful for my own work.
Read others’ articles on Medium. The more you read, the more you can get a sense of what is currently important in the data community and potentially get inspiration that way. In the process you will identify gaps in the topics covered, and perhaps you can add some value by filling in these gaps.
Draw inspiration from your daily interactions with people. It’s generally helpful to spend some time reflecting on your day. A lot of the inspiration for my articles comes from my interactions with people. When a lot of people ask you the same question, it indicates a need and a gap that you can help fill. I wrote an article about the best ways to use data visualization tools, because when onboarding team members, everyone seemed to be puzzled by what logic should go into data visualization tools versus data warehouses. Be a good listener and pay attention to people’s needs: those are great topics for your articles.
Channel your frustration. It doesn’t matter how perfect your job is, you will get frustrated about things during work. Use them to your advantage. You get frustrated about situations that were not perfectly handled. So ask yourself, “what’s a better way of handling this type of situation?” and “can I generalize this into a concept that can help others?” Because if you are running into this type of problem, it’s very likely others are experiencing the same thing, and your article can potentially help them.
As for finding time to write, the most important thing is to get over the mental hurdle. A lot of people think they don’t have time to write because they are picturing writing a school essay with a deadline, for which you have to spend a whole evening forcing things onto a piece of paper. In reality, you don’t need to sit down for 1–2 hours to write an article in one sitting. On the contrary, A lot of my good ideas come to me on my way to a coffee shop, or when I’m eating lunch. Take a minute to quickly jot them down on your phone. Write a little bit every day. It can be an intro, or even just a sentence: it doesn’t matter. The most important thing is to build the habit of writing.
Tessa Xie reflects on the twists and turns of data science careers, the benefits of consulting, and the value of writing for a wide audience.
In the Author Spotlight series, TDS Editors chat with members of our community about their career path in data science, their writing, and their sources of inspiration. Today, we’re thrilled to share our conversation with Tessa Xie.
Tessa started her career in finance as a quantitative researcher after graduating from MIT with a degree in financial engineering. She later worked in data science consulting at McKinsey, followed by a transition into the tech industry as a data science manager at Cruise, an autonomous driving company. Most recently, she started her new journey as a data science manager at LinkedIn.
How did you decide to become a data scientist?
I first became interested in data science in graduate school, where I majored in financial mathematics. Because there is a lot of overlap between financial mathematics and data science, we had the chance to take data science classes as electives for the curriculum. To my surprise, I found that I was more interested in the data science classes than the ones that focused specifically on finance.
However, I didn’t immediately enter data science after graduating, because I wanted to try out the industry I originally set out to explore — finance. After being a quantitative analyst for over a year, my initial hypothesis was confirmed: I like working with data, just not necessarily finance data.
I knew I wanted to leave finance but was not sure which industry I wanted to enter. Consulting seemed to be the best way to explore data science in different industries before settling into one. Consulting firms also interview data scientists from more diverse backgrounds than, say, tech, so it was a good opportunity to transition from finance to data science. Even though there’s overlap between quantitative finance and data science, I still decided to take a couple of online classes to refresh my data science knowledge, and those definitely paid off in interviews.
You recently spent some time in the autonomous vehicle sector. What is it like to work as a data scientist in that industry?
Because autonomous driving is such a futuristic field, one of the biggest challenges is the fact that we are solving a lot of data science problems for the first time. It’s a challenge because when you are thrown into a problem, more often than not you don’t have playbooks to learn from; you don’t have best practices that you can look up. It’s also an opportunity, because this means that as a data scientist you can take a lot of initiative, and it’s a great learning experience—not to mention that you will be able to drive meaningful impact even as a relatively junior data scientist.
Another challenge that I have noticed is that there is a huge amount of data being collected across different channels (e.g. lidar, radar, camera, etc.) for traffic or road-related information. Being able to efficiently and accurately connect data from those disjointed channels is crucial, since this will serve as the foundation of any data science analyses built on top. It provides an opportunity for data scientists to work super closely with data engineers and data infra teams. In the process, I was able to learn about data engineering and data infrastructure and appreciate the complexity involved in these areas.
The third point that’s worth mentioning is that since most autonomous driving companies are either still in the research and development stage or just started to commercialize their product, there’s not much customer data that you can work with as a data scientist. So if you are a data scientist who’s very interested in product analytics, user research, and/or experimentation, just know that you won’t have a huge amount of data to drive insights from (yet).
Do you have any other pointers for data scientists who are considering this industry?
When it comes to evaluating opportunities, I currently put two factors above others — learning opportunities and my interest and passion for the product. I was initially attracted to the autonomous driving field for exactly these two reasons. Because of the fact that it’s such a novel field, I knew there would be a lot of challenges when it comes to data. In my opinion, most challenges are opportunities in disguise; it’s the best way to force you to step out of your comfort zone and learn about new things. So I was very excited to enter a field that nobody has fully figured out and participate in the “zero to one” transformation.
Also, autonomous driving vehicles are a product that I have faith in and am passionate about. As a person who doesn’t know how to drive, I’m very excited about the possibility of autonomous vehicles and their potential. Not to mention that AV technology will hopefully be safer than human drivers and potentially save tons of lives we lose in car accidents every year. Being able to work on a technology that brings positive impact to the world and that I will be using one day made the work that much more interesting and meaningful.
You recently wrote about the experience of leaving a previous role as a McKinsey consultant. What questions should data scientists ask when they make a decision about joining or leaving a company?
Data science is a broad term that encompasses many different roles across a large number of industries, and it can be difficult for outsiders to get a good overview. To help with this, I have written a couple of articles about how to decide what type of companies you want to join and how to decide on the right role in the data world.
When deciding between companies, the biggest distinguishing factor in my experience is the size of the company. Joining a big corporation usually means you have more best practices to learn from and better, or at least more predictable, compensation, but at the same time less direct impact and potentially a more process-intensive corporate culture.
Smaller companies, on the other hand, usually provide more opportunities to take initiative and drive impact because a lot of things are not quite figured out yet. Contrary to the corporate culture, you will experience more of a hustle culture in smaller companies because things need to move fast. I went into detail about this in my article about whether you should choose to join a small startup or a big corporation. Of course there are outliers to any generalization, so you really need to judge companies case by case; but the type of questions you should ask yourself when making this decision should be more or less around the same dimensions — company culture, learning by doing or learning from best practices, compensation, and career outlook.
What about deciding which specific role to pursue?
When it comes to the type of role you want to choose, it’s a harder question because there are so many roles that can fall under the data science umbrella. There are data engineers, data scientists, data analysts, decision scientists, ML engineers, and more; it doesn’t help that a lot of companies use those titles interchangeably. But the most important question to ask yourself is: What do you like more, to explore problems or to implement solutions?
The more engineering components the role has (MLE, data engineer), the more it will be focusing on the “how” and you will spend most of your time on implementing solutions. The roles that are closer to business teams (data analyst, data scientist) will be focusing more on exploring the “what” and the “why.” As long as you know the answer to this, you should have an easier time mapping the job description to your interests.
Should these questions change depending on the career stage you’re at?
If you are in an entry-level position, these questions are less important—for the early years of your career, the most important thing is to learn. So the only thing that you really need to ask yourself is, “does this role provide me enough learning opportunities?” You will have plenty of opportunities to pivot your career if this role doesn’t fit all of your criteria. If you are in your mid-career, you will have an easier time answering the questions above based on your years of exploring and learning, and it becomes more important to focus on developing a clear profile that fits the career path you are interested in, as hiring managers and recruiters have increased the background requirements for experienced candidates.
You’ve written extensively about the interviewing experience for data science roles — why is it important for you to share these insights with a broader audience?
When I first decided to transition into data science, there were not as many resources online about how to prepare for DS interviews as there were for other roles, especially for people looking to switch from other types of roles and industries, such as finance. There’s the famous Leetcode for software engineers, and a book, Crack the PM Interview, for product managers. But when it comes to data science, I was more or less on my own for interview prep because it was such a new field.
After several rounds of recruiting, I have interviewed with countless companies, ranging from big tech to small startups and consulting firms. During this process, I have gotten a lot of insights from friends who are in this field, and have done a lot of research myself. After my job switch, I was getting a lot of outreach from aspiring data scientists through LinkedIn asking for my advice on interview prep. I’m big on supporting peers and avoiding repetitive work, so I thought it would be helpful for me to share my own interviewing experiences in a scalable way with a broader audience on a public platform like Medium.
I have recently started a new series specifically for interview topics covered in typical DS interviews. The goal is for readers to be able to use this series as a centralized starting point for their interview prep and job search. Speaking of which, I would love to get feedback from my audience on what would be helpful for them in their job search; I will try my best to incorporate those requests in future posts.
Finally, as a prolific author, do you have any advice for data professionals who might be interested in writing for a wide audience?
It’s a common thing for aspiring writers to suffer from imposter syndrome. I was definitely thinking, “I’m not a VP of Data Science nor am I a PhD in machine learning, how can I possibly add value to the data community?”
My advice is to forget about what you are not, and focus on what you are. It doesn’t matter who you are and what your experience is, you have something to bring to the table, and you can potentially help others with those insights. So finding that niche is key. If most of your work as a data professional is building models, I’m sure you have a lot of best practices you can share with people who are just starting out with modeling.
When it comes to finding new ideas for articles, I have three suggestions that I found helpful for my own work.
Read others’ articles on Medium. The more you read, the more you can get a sense of what is currently important in the data community and potentially get inspiration that way. In the process you will identify gaps in the topics covered, and perhaps you can add some value by filling in these gaps.
Draw inspiration from your daily interactions with people. It’s generally helpful to spend some time reflecting on your day. A lot of the inspiration for my articles comes from my interactions with people. When a lot of people ask you the same question, it indicates a need and a gap that you can help fill. I wrote an article about the best ways to use data visualization tools, because when onboarding team members, everyone seemed to be puzzled by what logic should go into data visualization tools versus data warehouses. Be a good listener and pay attention to people’s needs: those are great topics for your articles.
Channel your frustration. It doesn’t matter how perfect your job is, you will get frustrated about things during work. Use them to your advantage. You get frustrated about situations that were not perfectly handled. So ask yourself, “what’s a better way of handling this type of situation?” and “can I generalize this into a concept that can help others?” Because if you are running into this type of problem, it’s very likely others are experiencing the same thing, and your article can potentially help them.
As for finding time to write, the most important thing is to get over the mental hurdle. A lot of people think they don’t have time to write because they are picturing writing a school essay with a deadline, for which you have to spend a whole evening forcing things onto a piece of paper. In reality, you don’t need to sit down for 1–2 hours to write an article in one sitting. On the contrary, A lot of my good ideas come to me on my way to a coffee shop, or when I’m eating lunch. Take a minute to quickly jot them down on your phone. Write a little bit every day. It can be an intro, or even just a sentence: it doesn’t matter. The most important thing is to build the habit of writing.