A friend told me about a research scientist job opportunity, which aligned with my interests. I was sceptical about the place but tempted by their technological resources (Borg). The interviews were stretched over 2 months, mainly because of my availability. First, I had a short chat with HR about my work habits and motivation to join DM. The technical stage began with 2 rounds of mostly high-school level questions in maths, algorithms, statistics and ML – see Stage 1 qstns. The form of the interviews was confusing: a human person performing a quiz like a computer, “freezing” every time I tried to engage in a conversation. I prefer a conversational form, which gives an opportunity to explain my answers and gauge the strengths of potential collaborators. Next was a programming interview. Basic practical questions and a simple coding exercise in a shared Google document in a language of your choice. I use mainly C++, but since I heard that they dig Python, I read a tutorial, coded up some classics, and it felt more than enough – see Stage 2 qstns. The interviewer was a nice calm guy, we had a casual chat at the end. The last interview was with 3 members of the team to which I applied. HR gave me their names so I could read their publications. I was unpleasantly surprised: shamelessly naive research statements, an avalanche of arxiv papers reporting results of a simple AI model in different settings, claiming to address complex real-world problems, and Twitter sentiment analyses. The first guy was very late, didn’t apologise, seemed absent-minded and hardly interested. He decided that I must have heard about him so no need to introduce himself (he wasn’t anyone famous). Despite the disheartening atmosphere, I tried to talk about their publications, but judging from his blank face expression he didn’t recognise them. I mentioned a mathematical problem I saw in their method – he waved his hands and said that he knew what I was talking about. The next interviewer was also sure that I heard about him - I answered that I heard that they were all very nice. He wanted to sound ‘corpo’ and be very nice, but all he could muster was patronising. Looking at me like at a little puppy, he explained in a condescending voice that there was no place in DM for my pro-social research inclinations (highlighted in my cover letter), because "they are Google - they only want to make money!" Then he talked about how great and smart they were, and the great things they did. Suddenly, he asked about my career "changes". I know many people who made similar moves after the PhD/postdoc, but he was uncomfortably persistent about it. I had no spectacular answer to satisfy him - I followed my interests and good opportunities. Apart from that he asked me some short questions – see Stage 3 qstns. I had one more chat and attended just for the sake of the person who recommended me. I met an amusingly arrogant research engineer. Unusually, he took his time to introduce himself, even if, of course, I must have heard about him. He had a funny way of asking questions about my work, pretending that he knew what he was talking about. I worked in a spectrum of disciplines and problems, so I don't expect people to recognise them, but pretending that one does is a poor professional habit. He asked me to describe in detail the technique I used in my recent projects, which I “loved” according to him, and explain what I planned to do in DM. I sketched it in the simplest form (he asked how I draw events from a distribution rather than about statistical and code design problems). I decided to ask the person who recommended me about his interview impression – it was very distinct from mine. I wondered why: we have the same educational background (in fact my academic record was better, and I have more dedicated experience than he did when applying), similar skills, personalities and even similar poor sense of humour. There's one major difference: he's a man and I'm a woman. Sexism? Indeed, later I found that DM hire very few women. My CV was probably qualified by some diversity-aware AI algorithm (read: good old logistic regression), and those poor guys had to comply. Sometime later I got the decision from HR that they weren't interested. I thanked them, still gutted by my experience but with relief: sexism or not, after a closer look at those people and their research, “overrated” is the least I could say. For a fresh PhD or an aspiring academic, it’s probably a great job, which pays better than the uni and doesn’t require as much hard skills and effort as a regular money-making business, but the evident focus on quantity-over-quality and patting each other on the back to make a good impression on their sponsor/Google would be unbearable for me. Still, knowing a bit about bias in statistical research and how it can creep into algorithms and models, I hope that those guys will stick to Starcraft and not have much influence on my life.

7 -9 rounds depending on candidate. The rounds involved 1-3 scientists taking test on leetcode style questions, AI fundamentals, Transformers, Finetuning, Training approach of large models like Gemini, Deepseek etc.