The Training
What Amazon's hiring process actually looked like from the inside.
It was 2020. I was a Principal Engineer at Amazon, qualified to sit on hiring loops. The training was mandatory, something bland, the kind of title you forget. The instructor was someone from HR I had never seen before and would never see again. He had a lisp and an affect straight out of central casting, and he kept saying "la-tin-eks" with the careful enunciation of someone who had practiced it.
The training was about how to find and prioritize diversity candidates.
Not how to remove bias from your evaluation. Not how to assess technical skills without regard to background. How to find them. How to look at a LinkedIn profile and extract signals about whether someone was Black, Hispanic, or queer. HBCUs. Fraternities. Scholarships. The internal associations at other tech companies: the Black engineers' group, the Latino network, the LGBTQ employee resource group. Ethnic names, photos, rainbow flags in the bio.
The purpose was "prioritization." Prioritize outreach. Prioritize advancing them to interview loops. "Adjust" the hiring loop. "Adjust" the order of offers. The words were chosen carefully. Nobody said "prefer" or "discriminate." But we were being told, explicitly, to sort candidates by race, ethnicity, and sexual orientation before we ever evaluated their technical skills.
Managers were being held accountable for their diversity metrics. There were dashboards. I saw them: pipeline diversity, offer diversity, acceptance rates, all broken out by demographic, orgs ranked against each other. The dashboards were above my level, but they existed, and everyone knew they existed. The training was the implementation layer. This is how you hit the numbers.
There were about a dozen of us on the video call. L6 to L8, senior enough to be on hiring loops, senior enough to know what we were hearing. Nobody pushed back. Nobody asked questions. I sat there thinking: how can this be legal?
I didn't change how I evaluated candidates. I kept my bar where it was. But I watched the pipeline change around me. Weaker technical credentials kept getting advanced. Candidates I marked not-inclined kept getting overridden and hired. I counted at least three that year who got hired anyway. And the "non-diverse" candidates just stopped appearing on my loops. I don't know where they went. They were screened out somewhere upstream, before they ever reached me.
The instructor never explained why any of this was legal. He didn't offer a theory. Not affirmative action, not disparate impact remediation, nothing. It was simply assumed to be okay. This was 2020. This was what you did. The question of legality was not raised, because raising it would have been a statement about which team you were on.
I left Amazon with friction. Whether this was part of it, I don't know. I didn't exfiltrate the training materials. I didn't keep my notes. I'm writing this from memory, six years later, because I never stopped thinking about that VTC and the guy saying "la-tin-eks" while teaching us to sort humans by phenotype for a Fortune 5 company.
What The Law Says
Title VII of the Civil Rights Act of 1964 prohibits employment discrimination based on race, color, religion, sex, or national origin. This applies to hiring. It applies to the process that leads to hiring. It applies to the criteria used to advance candidates to interviews.
The training I attended was not ambiguous. We were instructed to identify candidates by race, ethnicity, and sexual orientation, and to use that information to prioritize them in the pipeline. This is not "removing bias." This is adding it, in a specific direction, with specific metrics, tracked on dashboards, enforced through manager accountability.
The legal theory that might justify this is narrow and does not apply. Affirmative action in employment is permissible only as a remedy for documented past discrimination, subject to strict scrutiny, and must be narrowly tailored. Amazon was not under a consent decree. There was no court order. This was company policy, implemented through HR, with no legal justification offered — because none was needed. Nobody was going to sue. The people being discriminated against didn't know it was happening. The people doing the discriminating thought they were the good guys.
The Machine
The instructor was not a rogue actor. He was not making this up. He was delivering a developed, legal department approved, training that had been developed, approved, and rolled out to every Amazonian qualified to sit on a hiring loop. The dashboards existed. The manager metics existed. The screening that removed non-diverse candidates before they reached my loops existed.
This was the machine. This is what "DEI" meant inside Amazon in 2020. Not unconscious bias training. Not expanding the top of the funnel. Explicit demographic sorting, taught in mandatory trainings, tracked in dashboards, enforced through performance pressure.
I don't know if it's still running, or if the training is still being delivered. But I sat through it, and I know what I heard, and I know that nobody in that room said a word.
This explains a lot about the AWS side of things. Maybe it’s turning around via purge. I’ve interviewed a fair number of ex-AWS engineers. They’ve been universally terrible. I’ve gone so far as tell the recruiting staff not to put them in front of me.
I wish I could remember all the details of a computer-based training I was required to take around 2019 at an international corporation based in Germany that took the employee through various non traditional sexual orientation scenarios and different (from America) cultural and racial perspectives, requiring us to click here and there as acknowledgement that we understand and/or agree. We tried every which way to bypass (what I considered) offensive content, but we had to sit through each part with active participation. Within that timeframe, we also saw a notable change in some previously quiet male employees sauntering through the office with full makeup and jewelry. If only they knew what people really thought.