Jaskaran Singh retweeted
How does Photokeeper achieve near instant image classification even on edge devices?
The trick is we are not using one single large image model, instead we are using a 5-step pipeline with each step scoring a single aspect of the image
A thread 🧵
My phone gallery is almost 60% screenshots, memes and random forwards from whatsapp. So I built an app that scans your photos and finds them for you (to delete)!
Introducing Photokeeper: extremely lightweight on device photo organizer tool. Fully open source. Runs on anything.
<10 ms per photo. 2.6 MB model. Runs fully offline, nothing leaves your device.
Jaskaran Singh retweeted
My phone gallery is almost 60% screenshots, memes and random forwards from whatsapp. So I built an app that scans your photos and finds them for you (to delete)!
Introducing Photokeeper: extremely lightweight on device photo organizer tool. Fully open source. Runs on anything.
<10 ms per photo. 2.6 MB model. Runs fully offline, nothing leaves your device.
Replying to @sabeer
To know that you are free to create your own meaning and happy to pursue it.
Jaskaran Singh retweeted
Replying to @AlfredoABrown
Imagine you're there watching the Wright Brothers get in their flying machine back in the early 1900s, and being like "in what scenario is this necessary?"
You're telling on yourself.
Jaskaran Singh retweeted
Arnab finally met someone he couldn't mute. 😂
Anmol walked into his studio and demanded proof for the 'Pakistan-backed CJP' claims. Arnab kept shouting, while the BJP spokesperson looked completely lost.
All that noise, not one straight answer.
Jaskaran Singh retweeted
Restaurants are dead.
Chefs are dead.
I just bought an airfryer and I can cook anything.
And recipes are on the internet for free!!
Replying to @paraschopra
Because taste changes with time.
Time series forecasting?
Jaskaran Singh retweeted
What is the purpose of higher education in the age of AI?
I had a very candid conversation with my students about pedagogy, universities, and what needs to change.
Sharing it here to hear what others think. The link for the full talk is in the first reply.
Jaskaran Singh retweeted
this is a bit different from what anthropic employees are posting on AI killing humans, but when I saw videos from the chinese humanoid games this year, I was terrified
eventually the armed forces of any country will soon have to face those humanoids equipped with aligned/unaligned superintelligence
just look at the rate of improvement, its only going to get better from here
In order to learn anything from scratch - you need to know how the universe was invented.
Start from abstraction and let your curiosity take the front seat.
respectfully fuck this list. you don't need to "master" virtual memory before you write your first kernel. you need to write a broken kernel, watch it segfault, and go find out why. that's where you actually learn what a page table is. checklists like this are how people stay "preparing" for two years and never ship a single line of CUDA
i stopped following advice like this. now i just start: naive kernel, profile it, get humbled, fix one thing, repeat. every os/systems concept on this list i actually understand, i learned by hitting it mid-optimization, not by reading it in a vacuum first. go write the broken kernel!!
Jaskaran Singh retweeted
every company is a rl environment
every company should build their own evals
every company should posttrain their own models
every company should have their own inference
every company should own their customer and enterprise data
the next decade of ai engineering will be about building this loop
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. 🔥













