Background: 5.5 years of experience, all of it at second-tier companies doing infra.
Goal: mainly looking to move into AI-related work. Target level is Google L5 equivalent or above. I'm not afraid of a heavy workload, I like healthy competition.
Timeline: started grinding problems and sending out resumes from early February, and by early June had gradually started getting offers and wrapped up interviewing.
Back in February the market still felt pretty mediocre. I mass-applied to a lot of places and most gave no response — only Amazon (the Stores org) got back to me. After the OA they scheduled an onsite. But I wasn't really interested in the actual work, and HR wouldn't let me interview for SDE3, so I lost interest.
In March I reached out to a former classmate for a referral. Interviews seemed to pick up around this time. After landing interviews with Adobe and Apple, I turned down the Amazon onsite.
Adobe: referral. Failed the first-round hiring manager interview — at the time I had no idea how to handle behavioral-question interviews and just rambled.
Apple: interviewed across five rounds total, for two different teams, both under AI/ML-adjacent infra orgs. Passed every interview, but the roles ended up getting filled internally. HR seemed a bit embarrassed about it afterward too. Overall Apple left a pretty good impression — the company holds itself to high standards on security and ethics, which I found appealing.
Microsoft:
Meta: got an HR reachout in March — both the SWE and PE recruiters contacted me. I specifically confirmed that SWE and PE interviews could run in parallel without affecting each other. HR told me Meta's E5 bar currently sits at 6 YOE, and recommended I wait until I hit that in late May before interviewing. I talked to HR again in early June, but by then I'd already accepted another offer, so I never actually interviewed.
ByteDance: got contacted about a few openings, including one where a manager reached out to me directly on LinkedIn. ByteDance and TikTok interviews can't run in parallel — you have to pick one. I picked a role I was interested in and interviewed all the way to the final round. Passed everything, but didn't take it. The process didn't feel very polished — for instance, they told me beforehand it would be a coding round, but it turned out to be a system design round. That said, the infra team was strong and the scope was decent.
Coupang: a manager reached out to me on LinkedIn (seriously, everyone should keep their LinkedIn well filled out — there really are managers who recruit directly off it). Their team's scope was really good and I really wanted the role. The manager was also great about it — waived the earlier rounds and put me straight onto the onsite. There really are a lot of Koreans there, but from what I saw everyone seemed decent, including the hiring manager, who was also quite good.
Onsite experience: the resume deep-dive went really deep, almost felt like a stress test — come prepared for that.
It also had a strong startup feel and seemed to have real fighting spirit, which I found appealing — the founder is still at the company. But the core business leans heavily on the Korean economy/e-commerce, which isn't ideal.
In the end I didn't get it — left with a lot of regret. I think the reason was that my scope wasn't strong enough.
Databricks: referral. The HR screen was a bit unpleasant, full of elitism. The phone screen was with the interviewer, who seemed strong and humble; I thought I did pretty well, maybe an 80/100. But they rejected me outright afterward. Kind of caught me off guard.
Nvidia: referral for an ML platform role. After two rounds I realized the team was actually doing ops work, which wasn't a match, so I didn't continue.
I also interviewed at a few other companies, like Figma, Mistral AI, and Resolve AI. Happy to write those up too if people are interested.
Update (2025-07-23): Microsoft — got referred into four or five roles and didn't get a single interview. Ridiculous. Felt like the postings were all just PERM ads.
Update (2025-08-05): Mistral AI interview.
First, the interview happens on your own computer — I'd recommend setting up your Python and pyenv environment ahead of time.
AI coding tools aren't restricted at all, so go ahead and get Cursor set up beforehand.
The gist of it: they give you a Mistral API token and have you build a small agent/RAG-style tool on the spot that calls the Mistral backend. This requires being very familiar with Mistral's API docs. I'd recommend feeding the API docs into Cursor ahead of time — that way one prompt gets you working code directly.
Discussion
Loading comments…