abhijat sharma

Computer Scientist working in Cryptography and Cybersecurity. Alumnus of @iiscbangalore and @iiit_hyderabad

[abhijat sharma @abhijat92]

@Amara__Don Your question answers itself, it's the view count! The guest doesn't matter for podcasters as much as these numbers (in millions)

[abhijat sharma @abhijat92]

@abhijeet_dipke If you are really so genuine, why have you spent the whole day with speculative counter arguments? Degree proof is immaterial, whether it's you or the PM, all anybody asks of a leader is authenticity

[soyuz @soyuzT9]

See you in Dehradun on 7th at 3 pm for an interesting discussion on "Boots don't Joke", the relevance of humour in the rough and tumble of the Armed Forces.

[abhijat sharma @abhijat92]

@kushal_mehra @6amiji @harshmadhusudan @and @Iyervval Congrats @kushal_mehra ! When does it release on amazon canada?

[Simon Kuestenmacher @simongerman600]

Here is a quickly updated map by @xruiztru to reflect the latest geopolitical development in Latin America.

[Simon Kuestenmacher @simongerman600]

This map shows the European colonial presence in North America as of 1750.

[Shashi Tharoor @ShashiTharoor]

Wishing the venerable Shri L.K. Advani a very happy 98th birthday! His unwavering commitment to public service, his modesty & decency, and his role in shaping the trajectory of modern India are indelible. A true statesman whose life of service has been exemplary. 🙏

[Aaj Tak Radio @aajtakradio]

तीन ताल में सुनिए @kamleshksingh ताऊ की 'बकवास' कविता

पूरा एपिसोड : https://t.co/depRNWv7oD

@kuldeepmishra
@007asifkhan

#teentaal #TTstaff #teentaalpodcast #aajtakradio #satiricalpodcast

[GDP @bookwormengr]

SOTA FOUNDATION MODEL BLUEPRINT - DARPA FOR INDIA:
************

I work in AI in leadership role for one the largest companies in the world.
Sharing my personal opinion, but based on facts & years of experience.

You haven’t seen everything yet:
—————————-
You heard DeepSeek.
Wait till you hear about Qwen (Alibaba), MiniMax, Kimi, DuoBao (ByteDance) all from China.
Within China, DeepSeek is not unique and their competition is close behind.
IMHO, China will have 10 labs comparable to OpenAI/Anthropic and another 50 tier 2 labs.
The world will discover them in coming weeks in awe & shock.

AI is not hard (I am not high)
————————————
Ignore Sam Altman.
Many teams that built foundation models are below 50 persons (e.g. Mixtral).
In AI, LLM science part is actual easy.
All these models are “Transformer Decoder only models”, an architecture that was invented in late 2017.
There are improvements since then (flash attention, ROPE, MOE, PPO/DPO/GRPO), but they are relatively minor, open source and easy to implement.

Since building foundation models is easy and Nvidia is there to help you (if not directly, then by sharing their software like “Megatron” that is assembly line to build AI models) there are so many foundation models built by Chinese labs as well as global labs.

It is machines that learn by themselves…if you give them data & compute. This is unlike writing operating system or database software. Also, everyone trains on same data: internet archives, books, github code for the first stage called “pre-training”.

What is part is hard then?
———————————-
It is the parallel & distributed computing to run AI training jobs across thousands of GPUs that is hard. DeepSeek did lot of innovation here to save on “flops” and network calls. They used an innovative architecture called Mixture of Experts and a new approach called GRPO. with verifiable rewards both of which are in open domain through 2024.

Also, there is lot of data curation needed particularly for “post training”
to teach model on proper style of answering (SFT/DPO) or to teach them learn to reason (GRPO with verifiable reward). STF/DPO is where “stealing” from existing models to save cost of manual labor may happen.

LLM building is nothing that Indian engineers living in India cannot pull off. Don’t worry about Indians who have left. There are plenty in the country as of today.

Then why India does not have SOTA foundation models?
———————
It is for the same reason India does not have Google or Facebook of its own.

You need to able to walk before you can run.

There is no protected market to practice your craft in early days. You will get replaced by American service providers as they are cheaper and better every single time. That is not the case with Chinese player. They have a protected market and leadership who treats this skillset as existential due to geopolitics.

So, even if Chinese models are not good in early days they will continue to get funding from their conglomerates as well as provincial governments. Darwinian competition ensures best rise to the top.

Recall DeepSeek took 2 years to get here without much revenue. They were funded by their parent. Also, most of their engineers are not PHDs.

There is nothing that engineers who built Ola/Swiggy/Flipkart cannot build. Remember these services are second to none when you compare them to their Bay Area counterparts. There is brilliant engineering to make them work at the price points at which they work.

Indian DARPA with 3B USD in funding over 3 years
———————-
What we need is a mentality that treats this skillset as existential. We need a national fund that will fund such 10 teams & the only expected output will be benchmark performance with benchmarks becoming harder every 6 months . No revenue needed to survive for first 3 years.

3B$ over 3 years will be a small change for 4T economy & GOI.

@dhume @protosphinx @vikramchandra @naval @balajis @AravSrinivas @bhash @TVMohandasPai @RajeevRC_X

[BCCI @BCCI]

What a moment this for the youngster!

A maiden Test 100 at the MCG, it does not get any better than this 👏👏

#TeamIndia #AUSvIND

[Vanishree Rao @vanishree_rao]

This is the beginning of something I’ve wanted for decades. To make a real dent. One deep enough that I can pour everything I’ve got into it. Deep enough for my daughter to be proud.

Today, we’re unveiling Fermah. Fermah itself might be less than a year old, but my journey toward it started 20 years ago when I first learned about zero-knowledge proofs. Their seemingly implausible nature, yet groundbreaking potential, hooked me. I spent a summer reading Neal Koblitz’s book on “A Course in Number Theory and Cryptography”, and I remember staring at the ceiling, trying to prove Fermat’s Little Theorem with whatever little I had learned by then.

After my undergrad in India, I had the incredible fortune to do my PhD in cryptography under the guidance of Prof. Amit Sahai. His foundational work in zero-knowledge proofs made me a fan before I even met him. When we first met, I called him “Sir” (like we’re taught in India to address our teachers), and it took him a whole week to convince me to stop! During that time, I worked on various projects that pushed the boundaries of zero-knowledge proofs, multiparty computation protocols, and more.

You know, growing up in India gives you this drive to survive and create value. I had that mindset too—always on the lookout for problems to turn into impactful products. Prof. Dan Boneh can attest to that! Over the years, I’ve taken many weak ideas to him, and he was kind enough to break each one down for me. Thankfully, he steered me away from all of them—except for the idea of a universal proof generation layer. That one had him excited, and he supported me in countless ways to start Fermah. I’m forever grateful for that.

So, why Fermah?

There’s one persistent headache for anyone trying to adopt ZK tech: spinning up cheap, performant, and reliable proving infrastructure. It's a massive challenge. So much so that L2s, for example, are stuck with either half-baked, unincentivized networks of provers or, at best, centralized networks. The complexity of managing proving infrastructure, paired with its “contextual” nature in contrast to the “core” goals of their business, has made it a weak link in any ZK project’s stack.

Specifically, sourcing prover machines, designing incentives that work, keeping those incentives balanced, maintaining liveness (without over-incentivizing), and keeping costs low (without under-incentivizing) is tough. It’s a whole project in itself.
That’s where a universal proof generation layer comes in. It can aggregate demand from multiple sources, ensuring high utilization of the hardware, which drives down the cost of proof generation.

It was clear to me this had to be built.

No better set of backers than @a16zcrypto
CSX Fund & @lemniscap! Also, truly humbled to be backed by our follow on investors@BanklessVC, @LongHashVC, @POpsTeam1, @publicworksfm, @zkv_xyz, @class_lambda, @daedalus_angels, @zero__dao, and @TheVelocityDAO - and others. The angels we have — @balajis, @musalbas, @nickwh8te, @sandeepnailwal, @jdkanani, @zac_aztec, @clairekart, @dlubarov, @sergey_nog, @gregoireljda -- these are people I have high admiration for. I mean… what a backing!

The core of it all is the team. I have managed to put together some crazy passionate low ego team that only cares to ship. This is exactly what you will see us do. Ship. And make a dent. Deep, significant dent.

Let's go!

[Kiren Rijiju @KirenRijiju]

Three cities, three Paralympics, three medals! Mariyappan Thangavelu does it again!
From Rio 2016 to Tokyo 2020 & now in Paris 2024 — another glorious medal in the Men's High Jump T63 winning Bronze Medal.
A legacy etched in pure determination.
The world witnesses your unstoppable spirit, Mariyappan.
Proud of you Champ!

#Cheer4Bharat

[abhijat sharma @abhijat92]

@hashtag_web3 BASQUE

[abhijat sharma @abhijat92]

@zkbankai @cryptobuilder_ @aztecnetwork @0xAbix very interesting and relevant to the zk ecosystem of today! :)

[StarkWare 🥷 @StarkWareLtd]

ZK is a public good that should belong to everyone.

Matter Labs, the corporate entity behind @zksync, has filed trademark applications in nine countries, claiming ZK as its exclusive intellectual property, even though they neither created nor contributed to the creation of this technology.

By using the legal system to claim a public good for itself, Matter Labs contradicts the ethos of crypto, Ethereum, and its own principles, which state: "We can make this world better by increasing people's freedom."

StarkWare, alongside other leading ZK companies and researchers, urges the community to demand that Matter Labs withdraw all their trademark attempts.

💠 Public Statement: https://t.co/2lYrvrJDoH

[Sunil Negi @sunilnegi669]

Brigadier Suyash Sharma,VSM visited the PCI today n presented his 3 books for PCI Library to Dada and to some members of PCI. Was an informal meeting.He also read some of the extracts of his books - Musings of Military Minds, Sense in the Non Sense, Two Hoots and Three Cheers

[Kapil Sharma @KapilSharmaK9]

Dear @IndiGo6E first you made us wait in the bus for 50 minz, and now your team is saying pilot is stuck in traffic, what ? Really ? we supposed to take off by 8 pm n it’s 9:20, still there is no pilot in cockpit, do you think these 180 passengers will fly in indigo again ? Never 👎 #indigo 6E 5149 #shameless

[Angelo Mathews @Angelo69Mathews]

I rest my case! Here you go you decide 😷😷

[abhijat sharma @abhijat92]

@HumansNoContext Litchi, dragonfruit

[Sportz Point @sportz_point]

A fairy tale last test for Stuart Broad

#Ashes2023 #ENGvsAUS