Q&A with an Aire Data Scientist — Frederik Durant
Frederik Durant has been a member of Aire’s Data Science team since September 2018. Here’s a quick snapshot on his work, and life beyond.
1. Coffee count so far today (16.11)
☕️/☕️ ☕️
Aire-time
2. As you approach your one year anniversary at Aire, what’s been your biggest accomplishment since joining us?
Since my arrival here, I have fully backed the ongoing effort to better integrate Aire’s Data Science function into the end-to-end DevOps pipeline. Starting from a business question, our data science experiments don’t end with a machine learning model that works in a Jupyter Notebook. Maximal automation of our credit risk model evaluations on performance and non-bias criteria, plus continuous integration in a client-facing risk assessment service, are key to receiving fast feedback for the next iteration.
3. What uncomfortable forest* are you glad to have taken a walk in since arriving at Aire?
* At Aire, we’re not afraid to take the hard path. Even if it takes more time or creates more pain. We want to ensure we do it the right way.
Having the DevOps approach to Data Science accepted in principle is one thing. Getting it effectively prioritised on the various team backlogs is another trade-off. When I joined Aire, our CTO Tim instructed me to “question everything”. So that’s what I did, as of my second week here. Not without proposing possible solutions, of course.
4. What are you most looking forward to getting your teeth stuck into next at Aire?
The credit risk models we produce have a real impact on people’s lives. That is the main reason why Aire operates in a strictly regulated environment. At all times, we must be able not only to explain the behaviour of these models, but also to trace back all the code and data that went into training them. This effort starts in the data science lab. My personal mission is to help further extend and run the necessary infrastructural and machine learning process foundations. This requires a careful blend of process control and the ability to stay flexible and agile. While not forgetting the human and organisational aspects, of course.
5. What do you think the biggest change to the Data Science team has been in the last few months?
Over the last 18 months, Aire’s Data Science team has doubled in size. At the same time, our approach is shifting from project-driven to product-driven. To support this growth, we’ve structured the Scrum teams along horizontal infrastructure and vertical product lines. Concretely, we’ve created multidisciplinary teams that bring together representatives from product, data science, engineering and operations. The key challenge is to maintain the right balance between stability and never-ending, fast-paced change.
Downtime
6. How do you stay up-to-date and keep your skills sharp?
Four years ago, I personally invested in a three month long Data Science bootcamp in New York City. Last year, I completed the “Deep Learning” and “Data Engineering on Google Cloud Platform” specialisations on Coursera. Aire also sponsors access to additional online learning resources. Furthermore, there’s an abundance of technical meet-ups here in London. One I attended was about Transfer Learning for Natural Language Processing, co-organised by my colleague Chris Howlin. Finally, at our own Lunch and Learn sessions, we discuss recent papers or exchange other novel insights and techniques, e.g. on Bayesian Networks.
7. What are you learning right now?
I recently finished reading Escaping the Build Trap by Melissa Perri and Inspired by Marty Cagan.
8. What’s been your favourite conference experience?
I really enjoyed the Strata Data conference last year. But Aire also hosts its own mini-conference called AireCon, where I presented a self-made visualisation of a British baby name similarity network for boys and girls. Other colleagues gave their own personal accounts on Jiu-Jitsu learnings, the need for waste reduction, and interactive fiction writing.
9. What’s the one conference you’d like to go to but haven’t been to yet?
It can hardly be called a conference, but I’d like to attend Burning Man some day. I guess there’s a hippie in me that hasn’t really come out yet. But for now, I’ll stick to PyData London, the conference we visit every year with our Data Science team.
10. What’s the podcasts/ blog/ book you swear by?
As a trained philologist I’m still keen on printed volumes. Here’s a list of professional books I find particularly valuable:
- Peter M. Senge: The Fifth Discipline. The Art and Practice of the Learning Organization
- Foster Provost, Tom Fawcett: Data Science for Business. What you need to know about data mining and data-analytic thinking
- Gene Kim, Jez Humble, Patrick Debois & John Willis: The DevOps Handbook; How to create world-class agility, reliability, and security in technology organizations
- Dave Gray & Thomas Vander Wal: The Connected Company
- Andrew McAfee & Erik Brynjolfsson: Machine, Platform, Crowd: Harnessing the Digital Revolution
- John P. Kotter: Leading Change
- Melissa Perri: Escaping the Build Trap: How Effective Product Management Creates Real Value
After reading, I also try to maintain my personal blog.
11. Any personal projects outside of work?
My weekly commute between Brussels and London does consume extra energy. So I’ve cut down on my weekend activities lately. Socially, I do spend a few hours a week at MundoLingo, to practice my language skills. Not in Python or R, but in French, German and Spanish.
