The Category Problem

My parents left Kanpur for Fremont in 1994, and for most of my childhood, caste was a word I associated with old-country baggage, something my parents' generation carried and mine was supposed to set down. I got a computer science degree, a job at a software company in the South Bay, a salary that would have sounded fictional to my grandfather, and for the first several years of my career I believed, genuinely, that I had landed somewhere the old categories didn't reach. I want to walk through why that belief was wrong, not as a confession, but because I think the mechanics of being wrong about it are more useful to lay out than the fact of having been wrong.

Start with a fact that surprised me when I first read it and still strikes me as underappreciated. A 2018 survey by Equality Labs found that roughly two out of three Dalits working in the American tech industry reported facing caste-based harassment on the job. Two out of three. That number sat in my head for a long time before I let myself apply it to my own experience, because applying it meant admitting that a series of things I'd filed under bad luck, or personality conflict, or just how competitive this industry is, might belong to a pattern with a name.

Here's the argument I want to make, and I'll build it the way I'd build a design doc, because that's the register I trust most and the one I think does this subject the most justice. The claim: caste discrimination in the Indian tech diaspora isn't a foreign import that occasionally flares up in isolated incidents. It's a structural feature of how hiring, mentorship, and advancement work inside a specific, tightly networked professional community, and it will keep operating invisibly for exactly as long as the institutions meant to catch discrimination don't have a category for it.

The mechanism is simpler than people expect once you actually look at it. Most senior engineering hires at companies with large Indian-American workforces flow through referral networks, alumni circles from specific Indian universities, regional associations, temple committees, WhatsApp groups built around shared hometowns or colleges. These networks aren't secret. They're just unremarked on, treated as the ordinary, harmless texture of a diaspora community taking care of its own. What almost nobody outside the community stops to ask is who these networks reliably exclude, and the honest answer, documented now in academic studies of Indian IT workplaces on two continents, is that they exclude Dalits about as reliably as the caste networks back home did, just rebuilt on foreign soil with a Silicon Valley zip code.

I didn't understand this mechanism the first time it affected me. Four years into my career, I watched two colleagues, both hired after me, both with less relevant experience on our specific stack, get fast-tracked into a mentorship program run informally by a director everyone called Sundar-ji, not his real name, warm, well-liked, from a Chennai Brahmin family, who ran what amounted to an unofficial pipeline for people he considered promising. I asked a mutual contact, carefully, what the actual criteria were for getting into Sundar-ji's orbit. The answer I got, delivered without any apparent awareness of what it implied, was that he tended to gravitate toward people who reminded him of himself at that age, people from similar backgrounds, similar colleges, similar, and I'm quoting this part exactly because the word choice matters, similar "sensibilities."

Sensibility is doing a lot of quiet work in that sentence. It's the modern replacement for a question nobody's allowed to ask directly anymore, and it produces exactly the same sorting the direct question used to produce, just with better legal cover and a cleaner conscience for everyone involved.

I want to address the strongest counterargument here, because I think this essay is weaker if I don't. Somebody reading this will point out that informal mentorship and cultural affinity exist in every industry, among every immigrant community, and that expecting a director to mentor people at random rather than people he clicks with is an unrealistic standard nobody actually meets. I take this seriously. Affinity bias is real and it isn't unique to caste. But the response to that argument is the same response Title VII eventually had to develop for race and gender in American workplaces more broadly: informal affinity networks that happen to sort along a protected characteristic aren't excused by the fact that the sorting felt natural to the people doing it. The law had to learn, slowly and through litigation, that "I just clicked with people who reminded me of myself" is not a defense when the people you click with all belong to the same historically dominant group and the people you don't click with all belong to the same historically excluded one. Caste is simply the category American law hasn't caught up to yet, and diaspora tech companies are, right now, in the exact position other industries were in fifty years ago, running an old exclusionary mechanism through new plausible-deniability language, protected almost entirely by the fact that nobody in HR has the vocabulary to name what they're looking at.

The 2020 Cisco lawsuit mattered for exactly this reason, regardless of its ultimate legal outcome. It was the first time a case forced an American regulatory body to ask, formally, on the record, whether caste belongs in the same conceptual family as the discrimination categories US civil rights law already recognizes. The lawsuit's specific facts aren't what I want to dwell on here. What matters is the structural admission embedded in the fact that the case could be filed at all: caste travels. It survives the ocean crossing. It reorganizes itself inside institutions that were built, explicitly, on the premise that this particular form of discrimination couldn't exist there, because the people writing the antidiscrimination policy had simply never had to think about it.

Seattle's 2023 ordinance adding caste as a protected category is the correct response to that admission, and I think every city and company with a substantial South Asian workforce should be treating it as the baseline rather than the exception. I've heard the objections to this, mostly from people in my parents' generation who worry, not without some historical basis, that formally naming caste in American law risks importing a stigmatized identity into a context where their children might otherwise have escaped it entirely. I understand the impulse. I think it's exactly backward. The alternative to naming the category isn't that caste stops operating. It's that caste keeps operating with no name attached to it, which is precisely the condition that let it survive successfully hidden for the first several decades of large-scale Indian tech migration to this country.

I left Sundar-ji's team eighteen months after that conversation, not dramatically, just a lateral move to a different group with a different director who, as far as I can tell, doesn't run any kind of informal pipeline at all, formal or otherwise, which has turned out to be its own quiet kind of improvement. I don't know if my former director understood what he was doing, in the sense of consciously excluding anyone. I suspect he didn't, and I think that's actually the more useful thing to understand about how this operates now, decades removed from anyone consciously enforcing untouchability with intent. The system doesn't need intent anymore. It just needs enough people quietly gravitating toward what feels familiar, at scale, across enough companies, for long enough, and the outcomes end up looking exactly like the outcomes intentional exclusion used to produce, just with plausible deniability built into every individual decision along the way.

This is the argument I'd want any HR department, any city council considering an ordinance like Seattle's, any well-meaning diversity officer at a company with a large South Asian workforce, to actually sit with. You cannot fix what your policy has no name for. You can run unconscious bias training every quarter, cover every protected category that Title VII already recognizes, and still leave an entire, well-documented, statistically significant pattern of exclusion running untouched underneath all of it, because the pattern operates through a category your framework was never built to see. Naming it isn't a symbolic gesture. It's the precondition for the audit, the complaint mechanism, the policy language, the entire apparatus that makes any other form of discrimination even theoretically addressable in an American workplace.

My parents didn't leave Kanpur so their son could spend his thirties making this particular argument in a country they crossed an ocean specifically to get away from these categories. I understand the irony. I've made my peace with the fact that the irony doesn't change whether the argument is correct, and I'd rather be the generation that names the mechanism clearly than the one that inherits it silently, the way every generation before mine was required to.