Principally Speaking: A Shapeshifting Journey in Tech
Navigating the curvy road that always goes somewhere in tech.
Benjamin Tan’s curiosity has always been strong, almost unstoppable, ever since he was a child.
The door to the computing world opened up to him when his parents brought home the household’s first computer when he was just an elementary schooler. While other children his age at that time tended to enjoy using the device, his interest was in how it actually works. His curiosity got the better of him, and he scrubbed the operating system, which led to his parents having to reinstall everything all over again. The occurrences happened so many times that his parents became semi-regular patrons of the repair shop, to the point where his father got irritated and told Ben to watch the repair person and ensure it wouldn’t happen again.
Still, they were supportive of his passion, despite not knowing whether Ben’s interest was just a phase or going to be a long-term thing. Gladly, their trust paid off as Ben stayed steadily strong, learning programming on his own before taking a bachelor’s degree in computer science at the National University of Singapore. He studied really hard during his time there and, afterward, was granted a scholarship to study technopreneurship at Stanford University in 2009. He managed to pave his way from Singapore to the bubbling tech scene at Silicon Valley.
Ben initially started out his career as a software engineer, doing web development for a French autonomous vehicle company. However, the seed of interest in machine learning has already been planted, and he grew it by learning from the ML team in his first company. There, he cultivated the skill in deep learning and built the necessary tools, which led him to eventually become part of the team himself. A new career path in his life is unlocked: becoming a Machine Learning Engineer.
In July 2021, Ben finally joined DKatalis as one of the first Machine Learning engineers, and as he spent more time at the company, his career expanded.
You’re one of the first members of our Data and Machine Learning team. Tell us about your journey in growing the team.
Benjamin Tan (BT): When I first joined DKatalis, the first year was spent really building out our machine learning operations capability because we started from scratch. So for the first three months, we started by building the Machine Learning platform. Then after that, getting the data scientists on board to use the tools because you need to teach them how to use them, you need to change mindset then as the team grew because when I came in, we had one machine learning engineer and two data scientists, which is kind of good because you can introduce things without a lot of destruction. Imagine if we already had 10 data scientists, then I would have to teach 10 people how to use it. In this case, I mean to credit how they were very supportive of my endeavors. Because They could easily say, “Ben, I don’t like the thing that you’re working on. I’m not going to use it.”
But I had their support, and we slowly started building the foundations, such as best practices, pipelines, and other important systems. This way, when other machine learning engineers join, we already have a pretty good set of things going, and they could come up with improvements and extensions. Now, we have grown by five times. And each person already knows the techniques and best practices. I think we’re in a pretty good state.
Once we have the machine learning infrastructure down, the next bit is how we make our work better and what tools we can use to do this. The ML Ops is always changing, with tools coming up left, right, and center. So, we spend some time doing proof-of-concepts to make sure that it’s our use case. We also embedded ourselves into the community quite a bit. So one of the things that I want to push people to do is to give back to the community, meaning, for example, in the Feast community, which is a feature store that we’re using, when we find bugs, we would highlight it, and at the same time, we contributed a patch. It’s a way of not only improving our work but giving back to the community.
Could you share with us what your team does?
BT: The data science team is really close to the business, and it’s slightly different from software engineering. For software engineers, there are multiple layers to go before you hit business. For us, for data scientists, it’s just one-degree separation. If you draw a line up, a data scientist will go directly up to a business person, the end of the story.
So, the projects that I’m handling would range across all the Business verticals. So, for example, I have projects that are doing fraud prevention and anti-money laundering, a lot of projects related to risk and go-to-market. And that’s just the business side.
There’s also the product side, where we would help people with onboarding, money storage, money movement, and so on. Our projects are not confined to just one aspect, they range across the organization. It’s good because the work is interesting and varied, but at the same time, there’s a lot of context to keep in the head.
You also recently became a Product Owner for Data Science, Risk, and Marketing Tech (MarTech), which brings more responsibilities to your plate. How are you embracing it so far?
BT: Well, it’s not like I’m playing two roles now. So I’m a product owner but also a principal engineer within the data team. It’s a lot more responsibilities.
I think the cool thing about being a product owner is that you see a lot more big-picture perspectives and the different dynamics that are at play. For example, being a product owner means that I can check with other product owners. Previously, when I was a machine learning engineer, all I could care about was data science, and that’s all. If someone told me about product strategies, I wouldn’t care much, just give me the fun coding projects.
But now I see products as a very important part of everything, and you can’t create products when you don’t understand the downstream. I think it’s one big opportunity that puts me in a really good position to ensure that the projects I’m working on make a high impact on the overall organization while also teasing out good projects for the data science team to work on.
Now, I’m obviously much closer not only with the product but also with engineering. I can already talk to the engineers in their language, and slowly learn to speak in the product language. There are some overlaps, but it still needs a fundamentally different set of skills. I’m taking this opportunity to really observe how the rest of the product owners operate and also try to get an idea of their thought processes.
How do you balance out all those responsibilities?
BT: The first thing is not to take on everything. If you have a good team, then obviously, you can always delegate tasks. But I think the other thing that’s very important is also to get regular feedback from multiple people regularly. So you’ll have to ask people, “Hey, how’s my driving like” every time because I mean, we’re all humans. A lot of times, you’ll have blind spots, and it’s always helpful to get feedback. Also, I guess one of the main things is to be humble.
One of the values that DKatalis believes in and pushes forward is purposeful growth. How do you put this into practice within your team?
BT: An interesting characteristic of the data science team is that everyone is highly educated. At that time, we had 3 PhDs and a bunch of Masters students. I kind of like to joke that I’m the least educationally qualified person on the team. Unlike in software engineering, where junior members might require more coaching and mentoring to upgrade their capabilities, a lack of skills is not a problem in the data team. When I recommend someone to take courses, it’s usually not for lack of qualifications.
What I’d like to do is direct them to good projects that can engage and require the team to use their skills to the maximum. Then, challenge them to think of creative ways where we can tinker with different use cases. We can go to business and say, “Hey, we have a suggestion for making this thing better using data science. What do you think?” And I think that’s a healthy dynamic.
In our last Retrospective, we discussed starting some kind of domain knowledge-sharing sessions, for example, in the realms of Go-to-Market (GTM) and other fields. I also think that when everyone is in the office, we can do reading groups. I’d also like for everyone to take different online courses and, after that, share the lessons with the rest of the team.
Talking about sharing, you have written 3 books and are an active Medium writer. What drives you to be productive and why is it important to share those articles?
BT: I believe that when you start writing things down, there must be clarity of thought. If you can’t think things clearly, you cannot put them on the paper or screen. Unclear thoughts will not work. Also, it’s a form of knowledge sharing, and teaching other people solidifies the concepts in your head. By writing books and articles, I not only get to share my knowledge but also clarify the thoughts and concepts I have in my head.
Furthermore, I write down the articles because they document my knowledge. There will be many situations where I need to get references; they were usually the same material I wrote some time ago. Every time, I was like, “Thank goodness I wrote everything down.”
Also, I believe that having the DKatalis engineers publish blogs will also help recruitment. As a software engineer, a company engineering blog will assist in assessing the level of engineering capability and quality. They can also get a glimpse of the projects that a particular company is working on.
Encouraging people, particularly data scientists, to build the blog writing habit is an uphill battle. They generally don’t like to write things down even though they are required to write documentation. But blog writing is different, and for some people, it’s a huge task mentally.
But let’s see, maybe I can bribe them with gummy bears or something.
Any tips for aspiring engineers?
BT: Never stop learning. I think software and data engineering are very unique jobs. The bar to learning is really low nowadays, as all you need is a laptop, an internet connection, determination, and grit. Many of the learning materials are now democratized, with people making high-quality code courses and content for free. There are so many high-quality resources available for free, so nothing’s really stopping you from digging in, understanding, and becoming an excellent software engineer.
Let me give you an example, if you want to be a dentist, you can’t go to Coursera or other online learning platforms and suddenly practice as one. For engineering, it’s completely different.
Interested in Ben’s helpful articles and tips for stepping up your Machine Learning game? Find them all here and don’t forget to hit the follow button!
