GitHub slashes engineering team in India
techcrunch.com
techcrunch.com
USA India
SDE2 252k - 325k | 65k - 100k
SDE3 400k - 600k | 150k - 200k
Also, Labor laws in India are non-existant where you can work your employee to death. Amazon also pays top of the market in India and you can get decent workers for much less. Hence, you can get 10x the output by investing the same capital in India instead of USA.
Another point is the labor of India being poor quality. These come because people contract WITCH companies which pay not even pennies to its employees. Expecting people to create top notch work while paying 4k$ per year is just not feasible even in India. (McD worker in US makes 13k$ per year).
It's been a while since I heard that term. They are everywhere in Danish government IT (through KMD).
InfoSys (the I in WITCH) invested in OpenAI back in 2015. How much is their investment worth now? If it is in the billions they should probably take the OpenAI hype and pivot to something else than being a cheap outsourcing company.
Nobody is claiming India is expensive when compared to the US.
They’re definitely not. Where are you getting this? Divide by three
Filtering on UK gives total comp of $60k-$200k+ at Amazon
So somewhere between you both.
The correct band for India should be around:
SDE1 - 30k
SDE2 - 40K - 45K
SDE3 - 50k - 70K
Then why doesn't every tech business around the world of any size do that?
It’s just they are outnumbered 1000:1 by some of the worst engineers I’ve encountered.
That set is hired entirely for low cost area-under-the-curve reasons. It’s sad because clearly there is interest in the profession (at least, seemingly, until their parents start asking why they aren’t VPs yet) but somehow companies and the educational system fail them.
For the most part India “worked” because it was so cheap you could hire an army of people and a crazy number of managers to make those people minimally productive. There is an art to that but it has enormous costs on the non-mediocre teams that have to burn tons of energy constantly course-correcting the remote teams and push back constantly on code quality issues.
I’ve worked for companies with massive india teams. Consistently they are pretty bad.
The companies who do it right are doing it to find talent, not cut costs, and ruthlessly filter and winnow the WITCH fodder until they get teams that are functional. But you have to be ruthless on quality and nepotism.
Median income in India is $3.6/day. It takes very little training to have someone do work for $10/day, which is a three-fold increase in income and a huge boost to standard-of-living. That's call center, data entry, etc.
From there, it takes a moderate amount of effort to train someone to churn out low-quality code. That brings income to perhaps $100/day, which is a very decent standard-of-living in a low cost-of-living area. That's maybe 2 years of bootcamp-level training and 4 years at a low-quality university.
Most of the talent pool in India (and many other developing countries) is shaped like a pyramid, where top-tier talent exists, but it's hard to find the needle in the haystack. Interviewing is far from a perfect process.
It's not a problem which is quickly solvable. The amount of top-tier talent can grow exponentially, but low-tier talent can grow just as quickly for many years -- India has over a billion people.
It's not clear it's even a bad thing. At the end of the day, the transitions from $1 => $10 => $100/day are a lot more important than $100 => $1000, since they provide for access to decent food, housing, and medicine. It's okay to prioritize growth at the bottom.
Without prioritizing the top at all, at some point, you do hit the middle income trap, but that's a much still better place to be than $3.6/day median income.
What has grown faster: the number of IIT seats or the number of "technical colleges" and bootcamps?
Manager no. 1 (HR) is a pushy, luxury SUV-driving, former IC who doesn't know anything about me or my work yet offers "advice".
Manager no. 2 (technical group) is a former IC who:
- Still plays at being a lead IC while also supposedly being an overworked manager
- Believes the world is orderly, the impact of work is entirely quantitative, and lacks theory of mind
- Believes he has a monopoly on good ideas
- Shoots down everyone else's ideas
- Refuses to concede any point or apologize in any context
Manager above them is a good guy, but all of his direct reports come across as a-holes.
I'm ready to bounce because:
- My coworkers don't value software engineering, curiosity, professional development, excellence, experience, aren't supportive in any sense, and not people I would socialize with... they're really boring. The ex team lead is a cool guy and one of the hardware integration people is my kind of properly socialized, hilarious troublemaker.
- On-call preventable fires happen often, sometimes in the AM
- "Too overworked" to do anything right the first time, so tech debt accumulates and there's no incentive to fix anything
- "Too overworked" and yet there's no possibility of new hires
- No one cares about the tech or language we have to use. Customers complain about it
- Nothing fun happens at work. Ever
Understanding when you are getting push back from an Indian team can be very difficult.
I've worked with people in India for 20+ years. Understanding how to read pushback and identifying issues early are skills you learn pretty quickly, but it is a good point.
As I said, there are awesome engineers there, and the key is to shape your hiring to the actual environment.
The hierarchy issue is humongous, though, even once you are really familiar with it. For me it is often just a blind spot for people who grew up in the US. I have many issues with US culture and the idolizing of titles, celebrities and rich people, but India is so often next level on this topic, coupled with a kind of acceptability of abusive behavior by the people higher up in the hierarchy.
Staying on top of this is both necessary _and an obligation_. You have a responsibility to the people you hire.
That's because a lot of the Indian consulting firms will take pretty much anyone off the street, put them through a 1-week "boot camp", and throw them at a project with the title of Senior Engineer*. As you said, there are plenty of talented engineers, but they're obscured by the sea of warm bodies whose only function is to put billable hours on an invoice.
* I haven't seen this firsthand, but it's what I've been told by Americans who have done management stints out there. It certainly seems consistent with the quality of work from some of the cheap job shops.
How?
Many Indians only work at these 30k jobs because of the possibility that they can eventually parlay those years of experience to emigrate and get a 300k job in the US. At big tech companies, many join the Indian office with the implicit or sometimes explicit promise that the company will sponsor their L1 or H1B after a couple of promotions.
I see parallels to how medical school graduates treat residencies and fellowships. If becoming a doctor meant making fellowship money forever (often 1/5th to 1/10th of what a full doctor would make), not nearly as many quality people would go into medicine.
You can only pay them so little because they see themselves as still in training for the "real job" that is in the US. Don't confuse this for true arbitrage.
As opposed to... working for that salary because it's an amazing salary by Indian standards?
I feel like this is my current work experience in the US. I frequently work 65 hour weeks and overnight shifts while also having to work all day the next day. There are no laws protecting me from this and I don't get paid overtime for any of it. I've worked the equivalent of 12 extra weeks of full-time work in the past year with no extra pay or time off to compensate.
(I've been casually looking for a new job/I'm quitting my current one in a few weeks)
WITCH: Wipro, Infosys, TCS, Cognizant, HCL tech.
<https://analyticsindiamag.com/witch-vs-faang-saga-to-continu...>
Comparable to FAANG.
MSFT saw a way to cut costs and the lax regulatory environment and just went with it. GitHub also closed down their SF office, had some layoffs here too and went fully remote
Great plan
1. Independently pick work. 2. QA it, sort out code reviews and then finally add them to the release cycle. 3. Write documentation or write change logs for it.
While all of this can be automated, there would need to be manual verification at some point (to fix broken tests or update configuration values) and that would require an engineer who is aware of the involved domain. And to adjust for vacations, sickness and other events, multiple engineers would be needed. All of whom need to have experience working with that area of the codebase, for which hands-on experience would be hard to gain, since the models are doing most of the grunt work. And at this point the company would be paying for both the engineers and the operational cost of running those models.
The operational cost is many times smaller than that of an engineer, and with proper loops it can work tirelessly.
If anyone is convinced of the tech it's them.
Cutting headcount is a consequence of cuts which is a consequence of priorities, legal, under performance, politics, etc. The very idea that some idiot has the power within the org to just sack developers because they have misunderstood how generative AI works is..... oh wait, that is extremely realistic because humans are very stupid. But also I would suggest it is not indicative of any sort of incoming trend and its unlikely to be best practice.
It's funny how even various "must haves" from the start of the project get cut, other things sneak in.
The back log grows and grows.
On the other side: if we increase productivity by say 3x, will we cut the backlog or will it grow 3x faster ?
I think we would, at least have a much better chance of exceeding our MVPs from the start of the project.
The reality is programming is a form of automation. The world's thirst for automation is insatiable. I don't think programming will always exist, but think of it this way: when you eliminate them, you automate automation. We've never been in a scenario like that, and it would be quite the thing to behold.
Think of it a bit like high frequency trading, but without the regulation or the barrier to entry. Sounds "great."
This assumes the company is making sound decisions. I see companies prioritise new sales over features that would make the software more usable and effective in existing deployments. This prioritisation of immediate cash flow makes it much harder for the organisation to understand why some customers ditch the product later or don't use it enough to knit it into their processes.
It is the perennial enterprise question as to whose needs are more important? The features the CEOs/CTOs of new customers ask for, or the productivity features the users of the products ask for.
In other cases operational efficiency can be easily backlogged just because its harder to perceive. I've seen specialists and consultants be required for installation and setup because key features required for ease of deployment and use are never prioritised. Its easier budgetarily to just write a bit of python to paper over some cracks than get the required features into the core product but this just broadens the code base and arguably increases the liability. Later, managers might hand-wring about why its so expensive to deploy and maintain without ever being able to prioritise the relevant tickets to address those issues.
Capital being spent on GitHub to develop it could be used elsewhere or banked as profit if you had a smaller team.
GitHub isn't in a position that it's market dominance will be threatened by a slower pace of development.
You're also talking in absolutes, GitHub says there's a 40% productivity boost from using ai. Assuming (simplistically) that this results in a 40% required headcount reduction for the same output, you can then see:
0% headcount reduction, 40% improvement in velocity
40% headcount reduction, 0% improvement in velocity
Scale these two factors mentally to see that it's not all or nothing, maybe they've gone 20/20 and so some people still get to keep their jobs.
That is their own marketing bumpfh and idk if any org should dog-food sales copy as strategy.
I am saying that cutting development teams by 40% on the wonky maths that generative AI makes developers 40% more "developery" won't be best-practice. I don't think its a leap before you look job and I don't think the toolsets and practices are mature enough to bank that 40%, at least, not yet. Therefore I think the slashing of the engineering team in India is not related to AI but rather budget or political or smth else.
I guess at the executive level its about whether you think you can increase sales with more R&D or whether you have entirely saturated your market or organisational capacity. Still, I would argue that mindlessly cost cutting is more the domain of the post-IPO, than that of the entrepreneur.
AI's impact will come, in next several years, but now it is too soon.
I thought indians were cheap (and that's why companies hire them in the first place). Is MS not able to pay them ?
Those firms are in the long run, condemned.
For me the move seems strange because a lot of companies in EU are expanding in India and are hiring a lot of indians. And, in contrast with China, they never said that the reason to open branches in India was to be able to get a foot on the market there. It's like India has cheap developers but no promising market.
On the same website, UK and Canada medians are way higher, a little under $100k.
You're saying it explains a multiplier of 4 between medians, do you have any source for that?
If I look for e.g. Apple ICT3 in India vs the UK it's $58k vs $132k.
https://www.levels.fyi/companies/apple/salaries/software-eng...
https://www.levels.fyi/companies/apple/salaries/software-eng...
That's still a multiplier of 2. For ICT5 the multiplier is even higher.
Amazon pays 160k$ to freshers in US. It pays 25L aka 30k$ in India. SDE2 band in India is 65k$ to 100k$. SDE3 Gets 150k$ to 200k$.
Add to that the lax regulations of labor laws in India and Amazon paying top of the market, they get a skilled and an experienced worker in India for the same price as a 21 yr old US college grad.
India is still extremely cheap compared to the west.
Government staff in india get ton of perks - cheap housing , very good pensions, healthcare for life and so on .
There is no easy way to cost this, that mission headline number is probably only the incremental mission specific expenses that got budgeted not the actual costs which are far higher if done via contractors (nasa) or private initiative (spaceX)
Both are way too high to believe.
as anyone who has actually been at a place that tried this knows, it's probably not true. and for the time when it is true, you need very good information on the talent, and experience on how to set up your product/engineering teams.
and then it still won't be as cheap as you expect.
Our Indian teams are making really good money these days, and if we don't pay them as such they bounce to another company.
If your argument is that there are more good devs in India by virtue of there being more devs/people there in general... well, sure, but that means there are more devs in general, and the ones making salaries better than the top 25% of Canadians are probably the top 1% in India.
Another friend avoided the reductions at Twitch last Friday thankfully but it seems the writing is on the wall. It is hard to make the decision to change job and lose out on a decent chunk of cash though. They've not made their mind up yet on what to do next.
This seems like standard corporate speak, but any conspiracist could also easily interpret it as the subtly-hinted communications of a mafia victim ... Are they being forced to do this by Wall Street? "Protect the health of our business" ... from vicious stock downgrades and short selling, "Grant us" ... future investment, etc
lol.
This shows that they just wanted to get rid of people where
1/ employment laws are weakest
2/ employees are least vocal
If they really wanted to save costs they'd be moving some open head count from US _to_ India. But that'd require them actually doing their job.
I'd be ashamed of myself if I were part of the team that approved this decision
Perhaps there's a startup idea here: identify the way communication is done with these teams (PRDs? I have no idea...), and create a service that takes that as input and creates results using LLM? There for sure should be room for fine-tuning models for it, as well as learning to have the LLMs create and run tests for all the code, according to requirements/specs.
Is FAANG laying off employees in Sweden as well?
Do you think most of the Swedish engineers working for FAANG would be replaced by GPT4 soon?
Are you planning of doing it in your startup as well?
Should the employees be worried?
As for FAANG layoffs - I think it is happening all over, but also that it is overreported. The graph of hires over the past few years should be references whenever number of people layed off is mentioned. Over the past 1-2 years I think the tech giants are way up in head count.
So many of these folks fill up seats in code factories or do glorified data entry.
I don't think the world is ready for what's coming.
The level of human interaction required to distill the abstract and implicit requirements of the customers into a machine-implementable format far outstrips the capabilities of current LLM based systems, and I say that as a daily GPT and Copilot user. It will just re-balance the workforce, turning 10x programmers into 50x programmer-project-managers, hence just ever more scarce.
I don't understand the math here: Why would it make programmers more scarce?
The whole 'AI taking programming jobs' reminds me of the '3rd world countries taking programming jobs'.
Don't get me wrong, its great for filling in parameters of functions and APIs, it sometimes good at figuring out errors and solutions. Its bad at coding anything other than generic algorithms, and its typically bad at that too.
That's pretty much what most of these firms do.
I typed in ChatGPT “write a better Windows” and have been copying code for days but it’s not really working out
-
Schreibe mir bitte ein Python script, dass 50% "I don't understand" oder andere 50% "I need a cup of tea" (nicht überserzt, nur auf English!) drucken. Alle Code Comments muss auf Hindi sein. Deine erklärung auser die Code muss ins französische schreiben sein.
~~~
```
# यहाँ हम शून्य से शुरुआत करते हैं
import random
# हम एक सूची में दो विकल्प डालेंगे
options = ["I don't understand", "I need a cup of tea"]
# हम 50% के लिए एक नंबर उत्पन्न करेंगे
random_number = random.random()
# यदि नंबर 0.5 से कम है, तो हम पहला विकल्प छापेंगे
# अन्यथा हम दूसरा विकल्प छापेंगे
if random_number < 0.5:
print("I don't understand")
else:
print("I need a cup of tea")
# यह स्क्रिप्ट आपको 50% की संभावना से "I don't understand" और अन्य 50% की संभावना से "I need a cup of tea" छापेगा।
# रैंडम संख्याओं का उपयोग करने के लिए हमने random मॉड्यूल का उपयोग किया है।
Voici un script Python qui imprimera 50% "I don't understand" et 50% "I need a cup of tea". Nous avons utilisé le module random pour générer un nombre aléatoire et imprimer une option en fonction de ce nombre. Les commentaires sont en hindi.
~~~~
(I know that code is trivial, that's not the point; I'm wondering how many requirements get lost in translation, both between languages and between jargon domains)
Also:
> # यहाँ हम शून्य से शुरुआत करते हैं
"Here we start from zero"
> import random
but then it imports a module from the standard library
FWIW, I also don't know about Francophone countries; sadly, despite the language lessons being mandatory for 5 years of my schooling, my grasp of it is so bad that even had to have my "I don't speak French" corrected.
Lemme guess: you generated that prompt by writing it in English and asking GPT (or Google translate, but not DeepL) for a German translation?
^_^;
(I assume "not DeepL" weil meine Übersetzung hat eine Irre, dass DeepL macht nicht?)