Are you tired of the same old tech recruitment process? Because I sure am.
The endless rounds of technical interviews and job postings that seem to prioritize random skills over broader abilities?
If so, you'll want to hear my story.
I am someone who has gone through the traditional tech hiring pipeline multiple times. But I was thrilled to discover a different approach when I applied for a job at the European Centre for Medium-Range Weather Forecasts (ECMWF).
In this blog post, I'll take you through my experience of how I landed my current job at ECMWF and contrast it with the typical tech recruitment pipeline. I'll share some of the unique aspects of ECMWF's recruitment process and the challenges and benefits of this process.
So, if you're curious about a different way of hiring machine learning scientists, read on!
I’m not a meteorologist. Barely a physicist, even.
My background is in geophysics, so basically a ground-wobble scientist.
But during my PhD, a lot of things went wrong, and a couple went right, which led me to work with a ton of machine learning related to various aspects of subsurface geoscience.
I wanted to move away from geoscience, which was heavily influenced by oil & gas and hoped to follow in the footsteps of Paige Bailey (@DynamicWebPaige), who pivoted from Chevron to Google.
Tech seemed like the place to be in our current economic climate, and I had a lot of applied machine learning experience that seemed to get me into some very interesting places.
But then there was a pandemic, and the world became complicated.
So I accepted a job that sounded quite fun, building up a new machine learning team in an established space and defence contractor with a few extremely capable colleagues. I quickly learned that the company was too set in their tracks to understand the impact machine learning could have if applied correctly and started applying to other jobs again. When my boss left for a significantly better salary, I knew the trajectory of our team was sealed.
Back to Linkedin!
Here’s my story of how I landed an amazing job at a place I want you to learn about, even if you’d rather work in tech!
The Common Tech Recruitment Pipeline
I have been through the tech recruitment pipeline multiple times, and I can tell you that it is a gruelling and exhausting process. Whether you're interviewing with a startup or a tech giant, the typical hiring process often involves multiple rounds of interviews, technical assessments, and take-home assignments. In this blog post, I'll draw from my experiences interviewing with Deliveroo and Amazon to illustrate some of the common elements of the tech recruitment pipeline and what I learned from them.
Amazon: An Entire Day of Interviews
The process was even more intense when I interviewed with Amazon for a software machine learning role. There's the usual pre-screening and technical phone interview. When I made it through, the company scheduled an entire day of interviews for me, including six separate interviews with different members of the organisation. The interviews covered various technical and behavioural topics and required me to stay sharp and focused throughout the day.
The Amazon interviews obviously needed even more preparation, as it was all in a single day. Additionally, Amazon has their core values that you should all know and align with. Finally, they were one of the few companies that had me do whiteboard coding, but it's likely I didn't encounter this in other places as interviews went online during the lockdown.
While the process was challenging, it also allowed me to meet a wide range of people within the company and get a better sense of what it would be like to work there. Yet again, many of the interviewers let me know that I would in fact never interface with them.
Why am I harping on about Amazon?
It is quite common for Amazon folks to go on and advise startups that try and adopt part of the core value stack but especially their interviewing pipeline. There are a few positives about Amazon’s pipeline. First of all, it seems like the recruiters and interviewers are “on your side”. It doesn’t seem adversarial, and the conversations were often very open and lovely. Obviously, the people also are very knowledgeable and they’re mostly aware of the ordeal you are being put through.
Concentrating and being social for an entire day without any proper breaks, however, is a reason I would assume that Amazon would filter out extremely capable people that otherwise fall under the neurodivergent spectrum. Socialising and concentrating for 7 hours with the spotlight on me just after I finished my PhD left me drained for weeks, unfortunately. Whiteboard interviews have been known to bias against under-represented minorities due to their test-like nature and attribution bias. It is nice to basically be done in a day though, that makes it easy to take time off work and prepare everything in a big chunk.
Deliveroo: Multiple Rounds of Interviews and a 24h Take-Home
When I interviewed with Deliveroo for a machine learning role, I went through seven rounds of interviews with different team members. The first round was an "informal chat" with a recruiter, followed by multiple technical interviews with engineers and managers. In addition, I was given a take-home assignment that required me to present a data analysis within 24 hours.
While the process was thorough, it was also time-consuming and stressful. Each interview required a significant amount of preparation and energy, and the take-home assignment was a considerable time commitment on top of that. Of course, while holding another job. However, the process did give me a sense of the company's values and culture. It allowed me to connect with different members of different teams, some of which let me know I would likely never interact with them.
I was impressed by the depth of knowledge and expertise among all the interviewers and sometimes even enjoyed the opportunity to showcase my skills in front of them. I often saw these people as peers I could have a lovely technical conversation with (maybe that was my downfall too, as I was quick to admit when I don't know things?)
Other Startups: Multiple Rounds up to the Founder
Generally, it seems common that you talk to a lot of people over a range of weeks.
The interviews are often with different experts with varying communication styles. So it can be somewhat difficult to know what to expect.
Here is a random assortment of questions I couldn’t answer because, quite frankly, they’re irrelevant to my day-to-day work:
- Multiply two matrixes by hand
- Where do the regularisation terms go in a loss function for neural networks?
- What is the loss in a Support Vector Machine called?
And I fondly remember the machine learning influencer on Twitter that proudly asked people in interviews to load a CSV without Pandas.
I encountered the most random tidbits of information that folks took as good indicators of machine learning prowess in interviews, assuming that if you didn’t know this one tiny thing, you couldn’t possibly know any other thing about a method. Unfortunately, this is quite common with folks that went through STEM education in specific institutions, where a specific type of thinking is the only way that will get you through. This, again is a reliable way to filter neurodivergent people out that often have a different approach to learning and understanding things. (Big scaffolding, then fill in the blanks, as opposed to small bricks building a house.)
When I applied to the ECMWF, I hadn’t even heard of them before. If you’re not a meteorologist, climatologist or part of NASA / ESA and similar institutions, like me, you likely haven’t either. In machine learning, we look at Google (or FAANG) as the big player in the space, hence why every Twitter bio that can will have “ex-Google” in there to give legitimacy.
I saw the ECMWF post on Linkedin. I was applying to every open machine learning position because it was a pandemic, and I knew my tenure at GMV would be short-lived after my boss was poached for an incredible salary.
The ECMWF website is fairly unassuming, to be honest, and I completely fell for it in the process.
It felt like an older, established place. No tech-glam. No foosball tables. No startup swag. Not even fancy animations on the website or fancy pop-ups to put in your email.
I sent in my application, I went through their pipeline (more on this in the next section), and I got an offer.
I got that job on the same day as another place, which made me quite the offer as well.
So I went to my old boss, and we had a conversation. I think he had to truly hold back from calling me an utter idiot and found the right words differently:
I have this friend who has been trying to apply for a position at the ECMWF for the past 7 years. Take the job.
I am a bit naive sometimes, admittedly. And sometimes, I also don’t quite realise the severity of situations or what I’m getting myself into. This was when I went back to the website and had a more thorough look at this place called ECMWF. Maybe the flags of the member states should have been a dead giveaway. Or the fact that it’s not just “using” but organising some of Copernicus. Or the fact that if you printed the staff regulations, you’d run out of paper and storing it on high shelves would be a health and safety hazard.
Should’ve been pretty clear indicators. But sometimes you need friends to see the forest for the trees.
So I backfilled all that information, learned more about this pretty incredible place and the implications of having a job here and possibly advancing machine learning in weather at what seemed to be a pretty influential place.
I probably shouldn’t say openly that it took me another chat to realise that the salaries posted were net of tax because of some partial diplomatic immunities. Pretty good surprise, though!
The Recruitment Process
After you got a pretty good idea of how chaotic I can be, you may be wondering how I stumbled into an opportunity like this.
The recruitment process is a little different to the classic tech pipeline.
The ECMWF relies on you doing some work upfront. You upload your CV, obviously, but most people get scared off when you have to fill out multiple text boxes detailing specifics of your experience and how it applies to the job posting.
However, I took this as an opportunity to explain how I’m changed from an oil & gas sponsored PhD to a space and defence contractor to now hopefully work in weather and climate. How my background in machine learning and physics would be great for this position and how my communication skills with a wide variety of stakeholders would benefit this position in the middle of ECMWF, between teams and disciplines.
It appears that this opportunity got me shortlisted to interview!
Now, I did read their outline for the recruitment process. But as interviews go, I figured there’d be several interviews during the interview stage.
So I was invited, we had a lovely panel discussion for an hour about machine learning in weather and how important it is to bridge the gap between domain scientists and machine learning practitioners. Don’t get me wrong. Getting that invite is intimidating. The only non-Doctor on the panel was the recruiter. And the outward perception (which ECMWF is trying to change actively) is that there are a lot of PhDs there.
(So just a note the position did and does not need a PhD! )
And then there was silence. I interviewed end of April. But I was used to companies going silent even if they proclaimed they gave feedback to candidates like the ECMWF writes on their pages.
Then I got a short message from one of the panellists if we could talk. I was offered a job in June! Apparently, there were some complications on their side, and then to top it all off I found the offer in my Spam folder as well. (That was the day I started checking my Gmail spam regularly…) So glad they went for the extra effort there reaching out to me.
Upon agreeing, the ball got rolling. They contacted my references, and I was sent to get a medical exam to be admitted to the private health insurance and the pension scheme.
End of June, I received my contract package, and two weeks later, I had my first day at the ECMWF on July 5th!
Was I surprised that there was just a single interview? Absolutely! Was I delighted? Even more so!
Just a very different experience from the typical tech pipeline I was exposed to for a while now. Turns out it paid off to put in a little extra effort with the questionnaire upfront.
The traditional tech recruitment pipeline can be a rough process that prioritizes narrow technical skills over broader abilities.
However, there are alternative approaches to hiring, such as the one offered by the European Centre for Medium-Range Weather Forecasts (ECMWF), which prioritize the whole person and their potential.
My story of landing an amazing job at ECMWF highlights the unique aspects of their recruitment process and the benefits and challenges that come with it.
Instead of looking for the next stepping stone that might make my Twitter bio a bit more interesting, it looks like I’m doing important work now that has a positive global impact.
I have to warn you, though.
I thought leaving oil & gas would finally get rid of the moral criticism. It does. But if you work at the ECMWF, all your friends will make you responsible for there being rain when the forecast only gave a 50% chance.