Interviews Interview with David Cournapeau, Head of the MLE Team: What qualities are required for Machine Learning Engineers?
In this fourth and final installment in our series of interview articles with David Cournapeau, he shares his experience working at Cogent Labs, where he heads the Machine Learning Engineering Team. David touches on the working environment at Cogent Labs, what motivates him, and what kinds of people he wants to work with in the Machine Learning Engineering Team.
Topic 3: The Joys and Challenges of Working at a Dynamic Company
Q: What are some of the challenges you have faced?
Before, I had only ever managed projects with three or four people. When you manage three or four people you can largely understand everything about the project, which means you can make decisions without so much ambiguity on the technical side.
Now, I have a team of many members. There simply is not enough time to understand everything anymore so you have to deal with ambiguity much more.
That is quite challenging, especially as I come from an engineering background. Engineers hate ambiguity and try to avoid it. At the same time, this is also more interesting than I thought it would be. I have to deal with too many things to fully understand them all and so I have to make many more judgment calls.
Q: How would you describe the working environment here?
DC: There are a number of factors that shape the working environment Cogent Labs. First, we are still a fairly small company so a lot of things are happening all the time.
This is to be expected, and fairly common in my experience.
Another aspect, which is more specific to Cogent Labs, at least in Japan, is that we are fairly international. We have a half-and-half mix of Japanese and non-Japanese members. Working out how to manage a large team of people with various backgrounds is a challenge, but one that I enjoy. In any company, different departments, such as sales and engineering, have their own “language” and “culture” so to speak. If you add in the different languages and cultures of people with different nationalities that can make communication even more complicated. At the same time, that is part of what makes it so interesting to work here.
Another factor that sets Cogent Labs apart and is actually one of the reasons why I joined the company is that our co-founders do not have technical backgrounds. Respectively, they worked in sales at Salesforce and in finance at Morgan Stanley. Before joining Cogent Labs, I almost exclusively worked for small companies that were founded by very technical people. Joining Cogent Labs gave me the opportunity to see another way of doing things. There are pluses and minuses. When you work with very technical founders, it is much easier to explain what you are doing, but it is also much easier for them to tell you how to do your job because they know a lot about it. At the same time, when working with non-technical founders, you may have more difficulty explaining the technical aspects of your work, so good communication is essential.
Of course, we work in AI, which is stimulating in and of itself. We are also using quite recent research, which makes our work even more challenging and interesting. Personally, I am more excited by how to developing products based on AI than the AI itself. If I wanted to work on the AI itself I would be a researcher. Dealing with the unique technical challenges of building AI-based products is what motivates me. This is still a new field so there is still so much that we need to work out, which is both liberating and sometimes scary.
Q: What qualities do you think would make someone a good fit for Cogent Labs?
DC: Let me answer specifically about the Machine Learning Engineering Team. For the engineers, I am looking for people who are software engineers first, but who are interested in and know about deep learning and machine learning. I want every machine learning engineer to really work on the product on a day-to-day basis. They need to believe that that is really what they want to contribute to. If they are not interested in working on a product and instead want to spend most of their time reading papers or training new algorithms, this is not the right role. We need people who understand that it is not just about writing software but making sure it works and making sure that it is not written in a way that only they can understand. The reality is that, even at a company like Cogent Labs where we try to apply rather advanced deep learning and machine learning, most of our work is still related to data management., maintaining accuracy under various conditions, scalability, etc.
Also, because we are still a small company, we can try to plan for the future, but realistically we cannot plan with much certainty nor plan too far into the future. Even if we try to set a specific direction, we know for certain that we will not follow it exactly. That means we cannot have people who are overly specialized in any one area. It is a tough balance to find but I want to look for people who are flexible enough to work in different domains – people who are more generalists than specialists.
Another important quality is humility. One of the things I have tried hard to promote is code reviews. When someone writes a piece of code, it is not simply pushed forward with the rest of the code. It must always go through review. I want people who are humble enough to accept that their code will always get reviewed.
For more senior people, I am looking for “T-shaped” people, to borrow an expression popularized by Valve, the video game company. That means people who have a certain breadth of skill and not just knowledge in one specific area. That breadth is the horizontal part of the “T.” At the same time, they also need to have deep knowledge in a particular field. That depth is the vertical part of the “T.”
Overall, rather than any specific skill, I am looking for people who already know something that is relevant to our company, are ready to learn something new as well, and are able to work in a team. If they have those qualities, they would already be among the top 20% of job applicants.