Pei-I / ML Engineer, Montreal

Q: What do you do at Jumio and what does that involve?

A: I’m a Machine Learning Engineer so I develop Machine Learning & Computer Vision approaches to solve business problems like helping to detect fraud or to anonymize our data. This involves many different things like data processing, reading academic publications to keep up to date with state of the art ideas and developments. Coming up with ideas so we can design frameworks and models to help detect fraud or generate images and other ML projects and then implementing them. The most interesting bit is coming up with new ideas after doing research into the latest developments.

Q: Before Jumio, did you have similar experiences at previous companies or was this fairly new to you when you started?

A: This is my first full time role in industry so it's hard to compare. Before joining I’d worked full time in an academic institution and done an internship but my experience was very different to what I have now. Here is much more excitement and energy about the varied projects and work at Jumio than in my previous environments.

Q: What do most people get wrong / misunderstand / not appreciate about the role?

A: People sometimes think ML is a magic wand that can solve any problem - when in reality it’s highly dependent on the input data. That’s why we spend a lot of time cleaning and processing data. Or they think of an ML model as a ‘black box’ that we just put data into and get results without knowing what’s happening inside. If you’re working on a computer vision project you can already easily visualize that each layer of the model learns. There are lots of little parts that build up the model and you need to know each part and what their function is. It’s like a recipe for how to compose your model in a way that makes it work. It’s true that there are still things we can’t fully explain why it works but the black box has been opened.

Q: What didn’t you expect about the role / what surprised you most?

A: Before, I thought I’d take on defined projects with concrete problems and definitions - but now in fraud detection we have to explore what we can do with ML and explore with the PM and my boss - we’ve no idea what ML can really do in this space so it takes cooperation with lots of communication to define what we really want to do and what we can do with ML.

It’s uncharted territory that we’re exploring together.

Q: What challenges did you face and what did you learn from that?

A: Remote cooperation has been a challenge - especially technical cooperation with other engineers in different locations around the world. When you can’t sketch ideas on paper face to face it’s sometimes more complicated to get those ideas across effectively.

Q: What would you say to potential employees wanting to work at Jumio? What qualities should they have?

A: Especially for this role you have to have a passion for learning new tools and technologies. It’s important to be able to adapt to changes as well as to appreciate others and your own skill.

Q: What’s the secret to being good at your job?

A: First you have to be excited about new technology that is just becoming popular. You have to have a lot of curiosity and want to know why something works or doesn’t work.

Just getting good results isn’t enough - you need to know why and how you got them.