Kelvin Yeung, Decision Lab UK’s Senior Consultant talks about his passion for bridging expertise areas that create new exciting solutions, what his parents think he does for a living (spoiler – drawing something on the computer), and what qualities a candidate must have to become a true Decision Labber?
What is your passion in Decision Lab?
Before I joined DL, I worked as an engineer in the railway industry. I did a lot of simulation modelling using AnyLogic. There is a lot of interest and the trend is for data science – machine learning and deep reinforcement learning. About a year ago, we realised that we can’t do just simulation or just data science. There needs to be an integration of the two.
I was recruited to Decision Lab to be a middleman to integrate the simulation side with the machine learning side of things so combined together, they can work as a single solution. It’s like a glue where you glue different parts together to a common single thing. This is what interests me the most.
What project you enjoyed the most at Decision Lab?
So far, I enjoyed pretty much every one of them. All of them for different reasons, as they all had different interesting elements to them. But there is one that I led called CIAO, which stands for compound intelligent agent optimisation. It is an AI agent that arbitrates several other AIs. It takes decisions and justifies which one is better. There is not anything quite like it around. I took time to research and developed this AI that is quite cool.
What exactly you enjoyed the most about that project?
It allowed me to use my creativity to develop something that didn’t previously exist. Like an inventor, I designed an architecture, not 100% sure whether it would work. I had to be flexible, research the things I wanted to try out and then test them as well. And fortunately, the client was happy [with the approach] as well. And everything went well, which was good.
Your dream client, your dream project?
I find the defence and aerospace ones the most interesting. I have been already working on the various defence projects, including naval ships. It would be good to work for somebody like European Space Agency [one of Decision Lab’s current clients], on rocketry stuff because I am a mechanical engineer.
Is it easy to explain to your family and friends what you do?
My parents ask: “What do you actually do? I never understand what you do.” I do find it hard to explain it sometimes. If I am doing simulation, most likely there will be a lot of things moving around and they’d think I’m just drawing something on the computer. It’s not too easy to explain, but I think they’ve got the idea now. I give them examples that help.
What industry or technical challenges intrigue you the most?
The defence sector is quite challenging because of the quality of work and the level of detail they require. In defence everything needs to be safe, so the requirements and the expectations are quite high. And that is also a reason why it is interesting.
What do you think a candidate must have to become a true Decision Labber?
The ability to self-initiate. If you just sit there and wait, no one might even know you exist. You have to be proactive and constantly communicate. Ask away, don’t wait for instructions.
The second thing, which is really a continuation of the first point, is whenever you run into a problem, you need to find someone and ask them to help.
The best bit about working at Decision Lab is…
All the people in DL are friendly. Even if we can’t physically see each other [due to the coronavirus situation], everybody tries so hard to make sure we are all happy and mentally healthy. People initiate lots of activities, such as Exercise Lab, Pub Quiz that they now host virtually.
Another thing to mention is you can use your creativity; you can speak out and people will encourage you and support you. There is an opportunity to pursue your passion and interest.
The next chapter is…
Continuing on the path of DRL [Deep Reinforcement Learning] and creating something new from several disciplines. DRL is the closest thing to an actual human-being. That’s why it’s so interesting for me and why I like it. There is a gap between the simulation models and AI. We should work on bridging this.