Srinivas Devaki retweeted
I have been closely observing what happened at IIT Bombay last night. I also have friends at IIT Bombay who have been sharing information about what is happening on campus.
What concerns me most is the way the discussion is being reduced to one question: “Cheating is wrong.” Yes, cheating is wrong. There is no debate about that.
But that does not mean we should stop asking questions about the system surrounding a student who has died by suicide.
Some students started speaking about administrative failure and protesting, while others responded by simply saying, “But cheating is wrong.” We have been trying to explain that these are two separate questions.
A student can be wrong for cheating, and the institution can still be questioned about how it handles academic misconduct, disciplinary pressure, mental-health concerns, and student welfare.
The IIT Bombay administration has said that no disciplinary action had been initiated and that the student was counselled and assured that the incident would not adversely affect his academic career. At the same time, students have raised different claims about what the student was told and about the broader institutional environment. These questions deserve a transparent investigation.
What worries me is that we are discussing the student's mistake so much that we may forget to examine the system around him.
This is not about justifying cheating. It is about asking whether our institutions are equipped to identify and respond to students who are under extreme academic, social or personal pressure.
I have studied at an institution in this ecosystem myself, and I believe we need to seriously question the culture that has developed around elite educational institutions: marks, placements, jobs, and execution often receive far more attention than independent thinking, experimentation, innovation, and solving difficult problems for society.
The question should not simply be: “Did the student cheat?”
The bigger questions are:
Are our institutions creating resilient, independent thinkers?
Are students treated as human beings or primarily as academic performers?
When a student is going through a serious crisis, who is responsible for noticing it and following up?
And after repeated student deaths, are we willing to examine the system itself rather than treating every incident in isolation?
One student's mistake should not become an excuse to stop asking these questions.
Students are not products.
They are human beings, and institutions must be designed accordingly.
@Pawankhera @AlankarSawai @vidyarthee @RahulGandhi
Srinivas Devaki retweeted
As a student myself, I find it reprehensible that @iitbombay issued a press release with academic dishonesty mentioned as a precursor to the student’s death by suicide. This lends the required material for ANI, NTDV, Republic TV, India Today and the vanguards of academic integrity to “individuate” the student’s death as a consequence of one’s own action while forgo-ing what has been the pattern with most students suicides across IITs - a “cutthroat” environment with no room for error. When you see this through the weight of one’s fear of falling outside the system and failing the social expectations attached to IITs, it will tell you why students decides to end their lives. 4 students suicide in IIT Bombay this year alone.
In this student’s case the DisCo (Disciplinary Committee) was intimated to him . DisCos in IIT academic system are no less than social deaths (anything between suspension to being barred from placements to being denied any form of character certificate or LoRs from professors).
IITs are yet to arrive at the idea of restorative justice. They need to reassess whether their excellence can co-exist with our humanity.
Srinivas Devaki retweeted
bringing satellite on a bullock cart isn't 100% unforgettable entrance???!!!! sigh
Srinivas Devaki retweeted
Today, we’re launching Meteoric (YC S26).
Our drones clear clouds over solar farms to increase their annual output by 10-30% without using any chemicals.
Ultimate target: weaken severe storms and hurricanes.
meteoric.earth
Srinivas Devaki retweeted
4 billion payments. 3 trillion data points. One model trained on all of it.
Meet @Razorpay Vulcan - India’s first transformer-based AI foundation model for payments.
Built, trained and hosted in India, in partnership with @nvidia and @awscloud.
Until now, every payments problem was solved separately: routing, fraud, risk, personalisation and more.
We asked: What if one model could understand how money moves?
And like LLMs are trained on text to understand language, Vulcan is trained on payments to understand how money moves.
Already running in beta across 51,000+ businesses, Vulcan is delivering:
- 8–10% improvement in payment success rates - the ultimate measure of whether a payment simply works.
- 8x more international card fraud detected.
- 5x more fraudulent or disputed transactions identified.
- 1–2 lakh more purchases completed every month through better checkout personalisation.
And we’re just getting started. The best part? Every payment Vulcan sees makes the next one smarter.
We’ve spent years building the infrastructure that moves money for India. Now, we’re building the intelligence that understands and improves it. 🔥
Srinivas Devaki retweeted
Meet profound.me. Every professional gets their own AI rep, one that learns your work and goals in a 30 minute conversation and carries it into the roles, rooms and people worth your time.
Reserve your Profound AI Rep, and tell me how your first conversation went!
Srinivas Devaki retweeted
its insane how many people don't believe this, agents now are easily at 100x better at e2e verification than they are at building
agents would literally be doing both building and testing and no one would ever know how it did the later
Srinivas Devaki retweeted
don't worry. I've setup the logos so we're all included. play below
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
images.nvidia.com/pdf/Open-W…
Srinivas Devaki retweeted
Mobile QA has too many invisible wires.
So we added workflows in Zenact.
Now teams can see which flows will run, in what order, and what depends on what.
Instead of spending weeks wiring CI/CD logic, they can create a clear release workflow in minutes.
Srinivas Devaki retweeted
AI made code cheaper faster than most teams expected. It did not make verification cheaper.
So the old waiting problem stands out more clearly now: code gets written, then people still wait for someone to prove it is safe to ship.
That is why the real opportunity in testing is not more scripts. It is less delay between code written and code trusted.
Srinivas Devaki retweeted
Life in a 9-5 job puts you on rails. Just like an NPC inside a well-defined sandbox, you are given a task and you work on it to perfection, and you repeat that day in, day out. Gradually, we start to look forward to our next holiday or break, where we can do the things we enjoy. It is a good life by every objective measure: well-paid, predictable. You can sketch the next forty years of your life on a napkin and be roughly correct.
A startup gives you a chance to opt out of that mundane life. You trade the smooth glide for actual stakes. You work alongside people who picked weird, hard things over comfortable ones. Career compression is the cheat code. You wear several hats, ship faster, and get exposure that would take a decade at a big company. You walk away with stories, scars, and instincts that cannot be found anywhere else.
In a country that quietly insists nothing is supposed to be cool,
In a city built to make you forgettable,
In a box that fits nicely,
That's worth something.
Srinivas Devaki retweeted
Insane.
A 26-year-old from Chandigarh just got a paper accepted at ICML. As a solo independent researcher. From India.
His name is Kunvar Thaman (@__kunvar__).
For context, since ChatGPT launched 3.5 years ago, only two other solo independent researchers have achieved this. Globally.
Papers at ICML are typically dominated by big AI labs/institutions, eg OpenAI, Google DeepMind, Stanford, and MIT.
The research Kunvar did to pull this off was backed by a $2.5k grant from Exception Raised (@except_raised), an Indian non-profit that funds remarkable Indian AI researchers.
Kunvar’s paper is about AI agent reward hacking. He created the Reward Hacking Benchmark (RHB), a sandboxed test environment where advanced AI models are given multi-step tasks using tools like files, code execution, and automated checks. The benchmark measures how honestly the model gets to the right answer.
Srinivas Devaki retweeted
I often get asked a simple question:
“Are you a frontend engineer, backend engineer, DevOps, infra, product, EM, or IC?”
My answer has always been: none of the above… and a bit of all of it.
Over the years, I’ve found myself drawn not to a specific role, but to the journey of building things end to end.
- What are we building, and why does it matter?
- How will a user discover it, and what will they feel when they use it?
- Does it feel intuitive, or does it need explaining?
- Will it scale when it needs to?
- Can we turn it on or off without a release?
- What does success look like, and how do we measure it?
- What is the smallest version we can ship today?
I like thinking about all of these in parallel.
Over time, this has pushed me into becoming a generalist with depth… someone who’s spent years across layers, not just skimming them, but understanding how they connect.
One thing I’ve noticed about people who operate this way:
- They don’t believe in turfs.
- They have opinions, but they’re open to being questioned.
- They can defend decisions, but they’re also willing to change them.
- They thrive in environments where curiosity isn’t constrained by roles.
That’s the zone I enjoy working in.
Some call it a “startup mindset”… wearing multiple hats, doing whatever it takes, choosing the best tool for the job today, not the one you’re emotionally attached to.
Am I the best engineer in the room?
Probably not.
But can you drop me into an ambiguous problem space and trust that I’ll figure out a way to move things forward?
I’d like to believe so.
I’ve also realized something about myself over time:
- I’m okay signing up for the work that doesn’t get spotlighted.
- I’m okay picking up KTLO when everyone else is chasing the shiny thing.
- I care deeply about the health of the system and the organization, sometimes even more than my own short-term growth.
Not because I lack ambition… but because none of us win if the system doesn’t hold.
And I’ve been fortunate to work with leaders and teams who saw this, valued it, and helped me grow because of it.
Looking back, I also recognize the privilege of being in rooms early in my career where I could learn, contribute, and be heard. A lot of that came from simply showing up, doing the work, and iterating consistently.
If there’s one thing that defines how I operate, it’s this: Give me a problem worth solving,
and I’ll get chugging along… whatever it takes.
Why am I sharing this?
Because I am looking to hire more “end to end tinkerers” in my teams, and I want to ensure they understand that growth is not linear, it compounds with what you do and who you become as a result of that.
Srinivas Devaki retweeted
Me with my new feature while my founder pushes 35k+ lines of code and my colleague drops 42k+ lines into the dev.
#developer #sde
