Leaving Amazon

OpenAI is so much fun!

I’d like to preface this by stating that Amazon is obviously a huge company, and my opinions are just that, one person’s opinions. There will probably be some people that share my frustrations while others have had a completely different experience.

I interviewed at AWS in early 2020, pre-pandemic. The interview process is grueling and I spent considerable effort preparing for the 5-hours-long pantomime of absurd algorithms trivia and “tell me a time when you said no” behavioral questions. COVID-induced visa processing delays pushed my start date forward in time many times. The high-stress interview process and years-spanning wait built up tremendous anticipation. In hindsight I can say I probably had somewhat unrealistic expectations when finally joining the company in late 2021.

Regardless, I was quite frankly shocked after my first couple of weeks, and my first impression was that this kind of work was not for me. As a software engineer, I expected to eventually do some software engineering. I’m not sure how to describe the work that first team I joined was doing, but I can’t in good conscious call it software engineering.

The bulk of the work consisted of updating configuration files and attending meetings with other teams where the driver would recite what was written in project management cards. No one seemed to really deeply know what the team was doing (or maybe I was too thick to understand it!), and quite frankly no one seemed to care very much. Having worked on very product-centric, fast-paced startups before, this kind of apathetic, complacent attitude was bizarre to me. If you browse Blind for more than five minutes, though, you will see for yourself that this kind of experience is far from being an exception.

I found my way out of that team as fast as I could. The ability to effortlessly change teams is something truly praiseworthy at Amazon. I got no pushback from my former manager. The whole process is extremely simple and streamlined.

Although the new team was far more interesting from an engineering perspective and my coworkers much more motivated, it was still an Amazon team. Over time I would understand that while Amazon has a great diversity of teams, they’re all playing the same tune. These are things that bothered me, in no particular order:

  • Big companies have a way of codifying their customs and culture into lists of commandments (1, 2, 3) that are often vague and contradictory. No one has figured out yet how to make true believers out of new hires, but you are expected to at least pay lip service to their moral code. In the day-to-day work life of rank-and-file employees, these dogmas are at best a hurdle that regular folk have to navigate around, and at worst become weapons to justify doing or not doing whatever you want.
  • Tooling. No matter how good an external solution is, there is an internal tool that is twice as difficult to use and half as good. This isn’t any developer’s fault, by the way, and almost always is a side effect of either legal requirements (software licensing), legacy decisions made long ago, or empire building. More on that later.
  • It is a large company where decisions flow from top to bottom. These decisions are to be accepted as facts of life. There is no conversation, no room for debate, and no one to appeal to. No exception, no accommodation. Scream against the wind all you want, write a petition with 30,000 signatures - doesn’t matter.
  • Big ships take a long time to turn around. Everything moves slowly and there is little space for experimentation. Do you have a good idea? Maybe something dozens of others also agree is a good idea? Good luck getting that into any roadmap.
  • Empire building. This follows almost as a direct result of the previous two points. Higher-level people decide things and those things get built, regardless of whether its a good idea or not. Promotions are vastly politics-driven, and large orgs seem to form around who has more political clout than any technical or business realities.
  • Apathy towards the craft. I struggled to find passionate, highly motivated and competent engineers at Amazon. I met a few excellent engineers, but most seemed to only punch in and out, with near-zero interest in anything related to programming or technology outside of working hours.
  • Misaligned incentives with the Leadership Principles. LPs are always the strongest force directing everything at work; engineers are trained to think about how their work artifacts align with the LPs, not on the quality or usefulness of what they are doing, which is usually left as an “if we have time for that” afterthought.
  • Unstable, untrustworthy leadership. I was truly blessed to work with first-grade direct managers. Higher level leadership, however, had some truly bizarre behavior the last few years, like the poorly-communicated, seemingly endless waves of layoffs or the disastrous return-to-office and return-to-hub policies.
  • Collaboration culture. While at any other job I would expect to be able to find a subject-matter expert on something, contact them and maybe do some quick pair programming, this is highly frowned upon at Amazon. There are good reasons for that: the company is too big for this kind of fast, informal interaction, and SMEs of highly demanded services would be quickly be overwhelmed by DMs if this were a thing. Moats are built to prevent that; instead of just asking someone about something, many times you will have to file a ticket and wait a week or more for office hours, which ends up being about as useful as DIY research in the internal platforms.

Some of the above is to be expected in any large company, and I could live with many of those issues. The one thing that convinced me that I needed to leave, though, was the stuck-in-quicksand feeling that working at Amazon was a career terminus. This was the first time in my career I felt that my day job was actively working against me, making me a worse software developer. There’s a choice to be made: keep up with the industry or keep up with Amazon’s internal technologies and tooling.

There was no new technology or innovative engineering process for me to learn. Simple problems become complicated ones because of the internal tooling overhead, leaving little time to work on more interesting things.

Going full circle, I want to restate the above is just my experience. I’m sure there are teams, maybe many of them, that contradict some or all of what I listed. I’ve never had demanding on-calls, for instance, which are an extremely common complaint at Amazon (I’ve personally seen friends and acquaintances leave a dinner party after being paged). To those who enjoy working there I wish only the best.